Service Level Agreements for Communication Networks: A Survey

AYYOUB AKBARI-MOGHANJOUGHI

JOSÉ ROBERTO DE ALMEIDA AMAZONAS

GERMÁN SANTOS-BOADA

JOSEP SOLÉ-PARETA

Department of Computer Architecture, Universitat Politècnica de Catalunya (BarcelonaTech), Barcelona, Spain.

Telecommunications and Control Engineering Department, School of Engineering, University of Saõ Paulo, Brazil.

[email protected], [email protected]

[email protected]

Abstract. Information and Communication Technology (ICT) is being provided to the variety of end-

users demands, thereby providing a better and improved management of services is crucial. Therefore,

Service Level Agreements (SLAs) are essential and play a key role to manage the provided services

among the network entities. This survey identiﬁes the state of the art covering concepts, approaches and

open problems of the SLAs establishment, deployment and management. This paper is organised in a

way that the reader can access a variety of proposed SLA methods and models addressed and provides an

overview of the SLA actors and elements. It also describes SLAs’ characteristics and objectives. SLAs’

existing methodologies are explained and categorised followed by the Service Quality Categories (SQD)

and Quality-Based Service Descriptions (QSD). SLA modelling and architectures are discussed, and

open research problems and future research directions are introduced. The establishment of a reliable,

safe and QoE-aware computer networking needs a group of services that goes beyond pure networking

services. Therefore, within the paper this broader set of services are taken into consideration and for

each Service Level Objective (SLO) the related services domains will be indicated. The purpose of this

survey is to identify existing research gaps in utilising SLA elements to develop a generic methodology,

considering all quality parameters beyond the Quality of Service (QoS) and what must or can be taken

into account to deﬁne, establish and deploy an SLA. This study is still an active research on how to

specify and develop an SLA to achieve the win-win agreements among all actors.

Keywords: Computer networks, Service level agreements, Service level objectives, Quality of experi-

ence, Quality of service

(Received May 19th, 2019 / Accepted June 1st, 2019)

1 Introduction

Over the years, the question of how to specify,

provide and measure service quality for network

end-users has been of utmost interest for service

providers, their clients as well as network infrastruc-

ture providers [88][43][98][114]. Furthermore, during

the last decade the liberalization and deregulation pro-

cess started in the telecommunication’s environment.

Increasing competition, favoured conjointly by client’s

performance needs, imposes huge pressures on service

providers and network. Moreover, for many years,

since having encountered particularly cost decreases,

these days, in order to differentiate their product from

the other competitors, providers try to improve quality

of service (QoS).

arXiv:2309.07272v1 [cs.NI] 13 Sep 2023

Hence, duty of each and every entity that partici-

pate in service provision and their relations need to be

explained. The domain is to elucidate liabilities of ev-

ery supplier and to assure quality of service needed by

client. The Service Level Agreement (SLA) is a beneﬁ-

cial tool in formalising the interrelationships among all

entities, that is the consequence of a negotiation among

all participating actors with the target of achieving a

common comprehension concerning delivery of ser-

vices, their priorities, quality, responsibilities, and other

relevant parameters [113][119]. To measure the agreed

SLA performance among entities, plenty of monitor-

ing tools and protocols have been developed by well-

known companies. As an instance, Cisco’s Service

Level Assurance Protocol (Cisco’s SLA Protocol) is a

protocol for performance evaluation which is deployed

extensively. This is utilised for service level parame-

ters measurement such as network delay variation, la-

tency, as well as frame and packet loss ratio. Further-

more, this protocol characterises the Cisco SLA Pro-

tocol Measurement-Type UDP-Measurement to enable

the interoperability of service providers.[20].

An SLA deﬁnes the guaranteed level of a ser-

vice property such as response-time, availability and

consequent actions in case of non-compliance situa-

tions, including liability as well as compensations is-

sues.

Based on the topological architecture of the net-

work, an SLA contract can be categorised into horizon-

tal and vertical SLA. Since the classiﬁcation of SLAs

in terms of horizontal and vertical categories depends

on the network topological architecture, which is out of

this article’s scope, both categories are introduced just

as follows. A horizontal SLA is an agreement between

two service-providers existing at the same architectural

layer (as an instance two Internet Protocol (IP) domains

[75][23] or two domains of Optical Transport Network

(OTN) [79]). On the other hand, a vertical SLA is an

agreement between two individual providers at two var-

ious architectural layers (for instance between an opti-

cal network and the core MPLS network). A network

topology diagram including potential SLA agreements

among different architectural layers is shown in Fig. 1

to illustrate the concept of horizontal and vertical SLAs

in a real scenario [74].

Other than the SLAs’ network topological architec-

ture, an end-to-end solution for management of SLA is

required to deﬁne services, parameters of Service Level

Speciﬁcations (SLS) and a classiﬁcation of the services.

The focus on the level of service instead of a network

An example of a paper based and a traditional SLA

contract is available at http://www.slatemplate.com/

ServiceLevelAgreementTemplate.pdf

Figure 1: A network topology including horizontal and vertical

SLAs, adapted from [74]

level enables the deﬁnition of SLA, services and/or

Quality of Service (QoS) independently from the under-

lying network’s technology. A service has to be deﬁned

without ambiguity utilising SLS and three information

types must be described: i) The QoX metrics as well

as their corresponding thresholds; ii) A method of ser-

vice performance measurement; iii) Service schedule.

QoX represents different quality requirements such as

QoS, Quality of Transmission (QoT), Grade of Service

(GoS), Quality of Resilience (QoR), Quality of Energy

(QoEn), Quality of Knowledge (QoK) and Quality of

Information (QoI) [93], and the mentioned quality pa-

rameters will be discussed in Section 3.

1.1 SLA actors and elements

1.1.1 Actors

A typical SLA involves two entities such as a contract

either between End-User (EU) and Service Provider

(SP), or between SP and Infrastructure Provider (InP).

In general terms, the complete scenario includes the

three mentioned actors. The term SP is referring

to corporations that supply data and communication

services to their customers. SPs may manage net-

works by themselves, or they may integrate the other

SPs services to deliver an entire service to their

clients/customers [93][1]. The SP can operate in differ-

ent business forms such as an Internet Service Provider

(ISP), a carrier, Application Service Provider (ASP) or

an operator. The client and/or customer are referring to

the end users that make use of all provided services via

SP. In this survey, we use EU instead of customers on-

wards. InP is another SLA actor which is rarely found

in the literature and mostly EU and SP are the actors

that are the focused point in the literature. The ISP

refers to the InP that provides physical resources, op-

erational infrastructure and the computing services for

development, deployment and management of the ap-

plications in enterprise class. For all involved actors in

an SLA negotiation, a win-win situation can be deﬁned

as quality requirements that are satisﬁed for all actors,

the EU is charged a fair price and the ISP and SP ade-

quately remunerated.

1.1.2 Elements

Table 1 lists the SLA key elements along with a short

description [72]. An SLA must be Speciﬁc and detailed

enough to deﬁne expectations for services and eliminate

any confusion. The Comprehensiveness is an essential

element of the agreement and the SLA contract must

cover all provided services by the SP and all possible

contractual obligations for all actors involved. More-

over, the SLA should be directly related to the service to

be offered and it must be Relevant to evaluating perfor-

mance against that goal. In the agreement, unrealistic

goals can demotivate the customers and non-delivery

will only lead to failures on agreed terms. Therefore,

the expectations set must be Realistic. By keeping the

language simple and Non-technical, for reference of

EUs, the contract would be easily understandable. The

responsibility should be clearly deﬁned as a set of Divi-

sion of work in the agreement. The SLA must contain a

Time frame against which the service will be delivered.

The Escalation Metrics must be clearly deﬁned. Once

the actor enters into the agreement, the client must be

aware whom to refer in case the services were not ren-

dered properly. Once all elements are considered in the

agreement, the agreement document must be the Au-

thoritative document binding all actors.

1.2 Literature classiﬁcation

To write this survey, more than 120 research articles

and technical reports have been reviewed. Based on

collected information on SLA, we deﬁned a compre-

hensive conceptual-map (see Appendix A) that served

as the basis for a structured-classiﬁcation of the SLA

literature. The literature on SLAs is in very various cat-

egories; organising and structuring the relevant works

in a systematic way is not a trivial task. The proposed

classiﬁcation scheme for SLA literature is illustrated in

Fig. 2.

We identify three categories: methodologies, mod-

elling and architectures for service level agreements.

This classiﬁcation allows grouping the reviewed arti-

cles under common umbrella and enables to review

three perspectives separately. Within the ﬁrst cate-

gory, methodologies for SLA, two subcategories have

Table 1: The best practices’ elements to draft a comprehensive and

reliable agreement

Element Description

Authoritative The SLA must be an authoritative document

binding both parties.

Comprehensive It should cover all services and all possible

contractual obligations.

Division of work The responsibility should be clearly deﬁned.

Escalation matrix The escalation matrix is a mapping of whom

should be contacted under a particular set of

conditions. For example, SLA must provide

an escalation matrix for handling any issues

of provided services by the SP and it must be

clearly deﬁned.

Measurable There must be a way to track actual perfor-

mance against the promised SLA.

Non-technical The SLA document must keep the language

simple for reference of non-technical people.

Quantiﬁed Deliverables should be quantiﬁed enabling

to be measured.

Realistic The expectations set by the SLA must be re-

alistic.

Relevant The SLA must be directly related to the ser-

vice to be offered and delivered.

Speciﬁc The SLA must be speciﬁc and detailed

enough to deﬁne expectations for services.

Time frames The SLA must contain a time frame against

which the service will be delivered.

been proposed: dynamic and static methods as well

as decision-making methodologies for SLAs. In the

second subcategory, two additional subcategories have

been identiﬁed: decision-making under uncertainty and

general decision-making methods.

In the second category of the proposed classiﬁca-

tion, we identiﬁed three subcategories: dynamic and

static modelling; application speciﬁc modelling; deci-

sion models for SLAs. Under the application speciﬁc

modelling subcategory, we can extend our discussion in

two different network classes: heterogeneous networks

and homogeneous networks. Using this classiﬁcation,

it is possible to accommodate different network tech-

nologies under these two groups.

In the architectures category for SLAs, we identi-

ﬁed four subcategories. The ﬁrst subcategory is SLA-

aware resource management and it has three subcate-

gories: virtual resource management, energy-efﬁcient

resource management and other resource management.

SLA monitoring and SLA management have been also

identiﬁed as subcategories for SLA architectures. Simi-

lar to SLA modelling we have application speciﬁc ar-

chitecture for service level agreements. This part of

proposed classiﬁcation focuses on speciﬁc architecture

Figure 2: Proposed classiﬁcation of SLA literature

proposed and/or deployed over heterogeneous networks

or homogeneous networks.

1.3 Paper contribution

This survey identiﬁes the current SLA research status

and provides a better understanding of existing meth-

ods and open problems in this research domain. The

identiﬁcation of research gaps and future research di-

rections points to how to utilise SLA elements to de-

velop a generic methodology for SLAs establishment,

deployment and management that will lead to a win-

win situation for all involved actors. Although there

are many surveys about SLAs, so far they are mostly

focused on particular services such as SLA in cloud

computing. This paper is organised in a way that the

reader can access a variety of proposed SLA methods

and models addressed to different applications domains

along with a detailed comparison. It needs to be em-

phasised that auction-based kind of SLAs are out of the

scope of this survey.

1.4 Paper organisation and reading-map

The remainder of this survey is organised as follows:

Section 2 introduces SLA characteristics and Service

Level Objectives (SLO). Section 3 discusses service

quality categories and quality-based service descrip-

tion. In that section we discuss four quality categories

namely performance, conﬁguration, data and security

along with their quality parameters. Section 4 reviews

the SLA methodologies proposed so far with focus on

their behaviour and decision-making strategies while

Section 5 reviews them from a modelling and architec-

ture perspective. Section 6 focuses on the open prob-

lems and future research directions. Finally Section 7

concludes the survey.

An “a la carte” approach can be followed to study

this survey and Table 2 provides a reading-guideline.

The readers with main interests in the detailed method-

ology and decision-making may focus their reading on

Sections 1, 4, 6 and 7. When it comes to those mainly

interested in the detailed SLA Modelling aspects, they

may read Sections 1, 2, 3, 5.1, 6 and 7. The readers

interested in SLA Architecture aspects can follow the

Sections 1, 2, 3, 5.2, 6 and 7. Finally, we recommend

Sections 1, 2, 3, 6 and 7 to the group of readers who are

interested in getting a very high-level overview of SLAs

including the characteristics and objectives of SLA as

well as open problems and future research directions in

this domain.

Table 2: A reading-guideline

Readers interested in SLA

Sections Basics Methods Models Architecture

1 ✓ ✓ ✓ ✓

2 ✓ X ✓ ✓

3 ✓ X ✓ ✓

4 X ✓ X X

5.1 X X ✓ X

5.2 X X X ✓

6 ✓ ✓ ✓ ✓

7 ✓ ✓ ✓ ✓

2 SLA Characteristics and Objectives

In this section SLA characteristics and a key element

named Service Level Objective (SLO) are explained.

SLOs are speciﬁc and measurable characteristics of

SLAs.

2.1 SLA Characteristics

The characteristics of SLAs can be identiﬁed as founda-

tion, change and governance characteristics [40]. Foun-

dation Characteristics (FC) of SLA includes provisions

which specify the key agreements and principles among

actors, the SP and their responsibilities and roles, and

the expected service performance levels [40]. The aim

behind the SLA agreements under FC is to publicise

the normal convictions shared by the two associations

with the goal that their Information and Communication

Technology (ICT) outsourcing relationship could con-

struct shared objectives and a general responsibility to-

ward the outsourcing relationship [21]. By characteris-

ing the goals and aim of the relationship, the objectives

that at ﬁrst drove the development of the relation can

be at least understood partially and shared by a group

of decision-makers and also individual employees who

inherit the relationship [21][56]. Moreover, clear proce-

dures of conduct can be set by these provisions through

deﬁning the responsibilities and roles of the diverse ac-

tors involved in the SLA.

Legally binding expression relates with SLAs

Change Characteristics (CC) including provisions con-

cerning processes for resolving prospective demands

unpredictable outcomes, processes for developing pre-

dicted contingencies as well as transformations, opera-

tions for recommending new innovations coordinated

with motivating force designs, and processes of efﬁ-

cient adjustments and feedback in the bilateral con-

tract [40]. The mentioned provisions are grouped un-

der the change characteristics of a bilateral agreement.

Moreover, these provisions over-attempt to make the

ground procedures and rules for dealing with prospec-

tive contingencies. These provisions are depend on

favourable outcomes as the environment of ICT devel-

ops quickly and the condition of business frequently

needs quick reaction from the providers to provide new

services or amend the running services [55]. Although

a comprehensive contracting is not a practical option,

because of limited rationality, limited but intentional

rationality is translated into incomplete but farsighted

contract. Indeed, previous studies in Information Tech-

nology (IT) outsourcing have shown the impacts of

evolving speciﬁcations and investigation the possibility

for extremely unstructured and/or uncertain tasks [21].

Contractual qualiﬁcations connected with Gover-

nance Characteristics (GC) of SLAs characterise the

ways to keep the relationships through clear measure-

ments statement, incentives and penalty, options for

termination and accountabilities, and well-deﬁned pro-

cesses for documented communication as well as pro-

cesses for resolving and recognising of potential dis-

putes [40]. In this way, the contractual bases underlying

governance speciﬁcations set administrative procedures

to continuous assess of the value which the correlation

is producing for the diverse stakeholders and to guaran-

tee that the relationship remains on course. Both ﬁnan-

cial analysts and authoritative scholars similarly sup-

port the spirit of the GC. For instance, it is proposed that

rewards and results be connected so as to prevail with

regards to overseeing between hierarchical connections

(e.g., [55]).

2.2 Service Level Objectives (SLO)

An SLO is a key element of SLA among its entities.

SLOs are agreed as a means of performance measur-

ing of the provider and simultaneously they are out-

lined as a way to avoid disputes between the two ac-

tors based on misunderstanding. The establishment of

a reliable, safe and QoE-aware computer networking re-

quires a set of services that goes beyond pure network-

ing services. Therefore, in the paper this broader set of

services will be taken into account and for each SLO

the related services domains will be indicated. There

is often times confusion in the use of SLO and SLA.

SLOs are particular and measurable characteristics of

the SLA like throughput, response time, availability, or

quality [122]. Apart of SLO application domain, we

can categorise the objectives in four categories, namely:

i) performance service level objectives; ii) security ser-

vice level objectives; iii) data management service level

objectives; iv) personal data protection service level ob-

jectives [104][11].

2.2.1 Performance service level objectives

The performance SLOs categories are given in Table 3.

For each category of performance SLO, relevant SLOs

are given with their description. This table covers i)

Availability, ii) Response time, iii) Capacity, iv) Capa-

bility indicators, v) Support and vi) Reversibility SLO

categories.

2.2.2 Security service level objectives

Security SLOs can be classiﬁed into eight SLO cat-

egories named: i) Reliability, ii) Authentication, iii)

Cryptography, iv) Security, v) Logging, vi) Auditing,

vii) Vulnerability and viii) Service change. For each

security SLO category relevant SLOs are given and ex-

plained brieﬂy in Table 4.

2.2.3 Data management service level objectives

Data management is one set of SLOs categories that

covers data relevant categories. As an instance, data

classiﬁcation deﬁnes two service oriented relevant

SLOs. Table 5 also addresses the other SLO categories:

data mirroring, backup, life-cycle and its portability.

2.2.4 Personal data protection service level objec-

tives

Data protection and especially personal data protection

are speciﬁc and measurable categories of SLO. Table 6

addresses eight individual SLO categories along with

their relevant SLOs.

Table 3: Performance service level objectives

SLO Category Relevant SLO Service Domain Description

Availability

Uptime level (availability) Connectivity, Cloud Comput-

ing, VPN

, SDN

, SFC

Shows the availability of service a certain period of the time,

over the aggregated feasible available time (in percentage).

Successful requests percent-

age

Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Indicates the number of error-free requests processed by the

service upon the collected number of submitted requests (in

percentage).

Timely provisioning service

requests rate

Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Indicates the number of provisioning service requests accom-

plished in a certain period of time over the total number of

provisioning requests service (in percentage).

Response time

Average/Mean response time Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

The statistical-mean upon a set of observed response time ser-

vice set for a particular type of request.

Maximum response-time Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

The maximum response-time goal for a particular and spe-

ciﬁc type of request.

Capacity

Maximum resource capacity Connectivity, Cloud Comput-

ing, SDN, SFC

The highest and available amount of an allocated resource to

an instance of the service for an special service client. Ex-

ample of resource includes data storage, memory, number of

CPU cores, and etc.

Simultaneous service users

number

Connectivity, Cloud Comput-

ing, SDN, SFC

The maximum number of individuals that can be utilising the

service at the same time.

Simultaneous connections

number

Connectivity, Cloud Comput-

ing, SDN, SFC

The maximum number of individual connections to the ser-

vice at one time.

Service throughput Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

The minimum number of speciﬁc requests that can be pro-

cessed by the service in an offered period of time like the

number of requests per minute.

Capability indicators

External connectivity Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Explains the service capabilities to connect to external sys-

tems and/or services.

Support

Support hours Cloud Computing, VPN Indicates the period of time in hours that SP provides an

EU support interface that allows the requests and general in-

quiries from the service client.

Support responsiveness Computer Networking, Cloud

Computing

The maximum period of time that the SP will take to acknowl-

edge a service client request or inquiry.

Resolution time Computer Networking, Cloud

Computing

The target resolution-time for client requests, in other words,

the time taken to accomplish any essential actions as a result

of the request.

Reversibility

Period of data retrieval Cloud Computing Indicates the period of time that the EU can retrieve a copy of

their service client’s data from the service.

Data retention period Connectivity, Cloud Comput-

ing

Indicates the period of time that the SP will maintain backup

copies of the service client’s data within the process of termi-

nation.

Residual data retention Connectivity, Cloud Comput-

ing

Elucidate a description of any data related to the service client

which is maintained since the termination process ﬁnished.

Table 4: Security service level objectives

SLO Category Relevant SLO Service Domain Description

Reliability

Service reliability Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Explains the service ability to accomplish its tasks accurately

without any failure over some determined period of time.

Redundancy level Connectivity, Cloud Comput-

ing, SFC

Describes the level of service supply chain redundancy.

Authentication

User authentication and level

of identity assurance

Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Represents the Level of Assurance (LoA) of the mechanism

that is used to authenticate an EU to get access to a resource.

Authentication Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Represents the available authentication-mechanisms sup-

ported by the Conﬁguration Service Provider (CSP) on the

offered services.

Revoke user access mean time Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

The average time required to cancel the users’ access to an

speciﬁc service on request over a particular duration.

Storage protection Computer Networking, Cloud

Computing, VPN

Refers to the mechanisms utilised to protect a service-user

access credentials.

3rd-party authentication Connectivity, Cloud Comput-

ing

Identiﬁes if 3rd-party authentication is supported by the ser-

vice and also it describes which technology/technologies can

be utilised for 3rd-party authentication.

Continued on next page

Virtual private network

Software-deﬁned networking

Service function chaining, also known as network service chaining

Table 4 – continued from previous page

SLO Category Relevant SLO Service Domain Description

Cryptography

Cryptographic brute force re-

sistance

Cloud Computing, VPN,

SDN, SFC

Expresses the cryptographic protection strength applied to a

certain resource based on its key length.

Key Access Control Policy

(KACP)

Cloud Computing, VPN,

SDN, SFC

Explains how a cryptographic key is protected strongly from

access, when it is utilised to come up with security to a par-

ticular service.

Level of cryptographic hard-

ware module protection

Cloud Computing, VPN,

SDN, SFC

Refers to the protection level which is given to cryptographic

processes through the use of cryptographic hardware modules

in the service .

Security

Timely incident reports rate Cloud Computing, VPN,

SDN, SFC

Refers to the outlined incidents to the services that are re-

ported to the clients in a convenient time (in percentage).

Timely incident responses rate Cloud Computing, VPN,

SDN, SFC

Indicates the outlined incidents which are acknowledged and

assessed by the SP in a convenient time (in percentage).

Timely incident resolutions

rate

Cloud Computing, VPN,

SDN, SFC

Explains the percentage of deﬁned incidents on the service

which are resolved after discovery and within a predeﬁned

time limit.

Logging

Logging parameters Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Indicates the parameters which are obtained in the log ﬁles of

a particular service.

Log access availability Cloud Computing, VPN,

SDN, SFC

Refers to the level of the service customer access to the log

ﬁle entries.

Logs retention period Cloud Computing, VPN,

SDN, SFC

Indicates the duration that logs are available for further anal-

ysis.

Auditing Certiﬁcations applicable Cloud Computing, VPN,

SDN, SFC

Indicates a group of certiﬁcations maintained for a particular

service by the SP.

Vulnerability

Timely vulnerability correc-

tions rate

Cloud Computing, VPN,

SDN, SFC

Indicates the number of vulnerability corrections performed

by the SP (in percentage).

Timely vulnerability reports

rate

Connectivity, Cloud Comput-

ing, VPN, SDN, SFC

Indicates the number of vulnerability reports generated the

SP to the service user (in percentage).

Vulnerability corrections re-

port

Cloud Computing, VPN,

SDN, SFC

Refers to the mechanism that the SP informs the customer of

vulnerability corrections applied to the SP’s systems.

Service change

Service change reporting Cloud Computing, VPN,

SDN, SFC

Refers to the sort of changes (such as functional changes or

SLA change), mechanisms and a certain duration for SP to

inform service clients of planned changes to the service.

Timely service change notiﬁ-

cations percentage

Cloud Computing, VPN,

SDN, SFC

The aggregate number of notiﬁcations for any change made

during a particular period of the time over the whole number

of notiﬁcations of change, in percentage.

Table 5: Data management service level objectives

SLO Category Relevant SLO Service Domain Description

Data-classiﬁcation

Service client data use by SP Cloud Computing, SDN, SFC Indicates the stated policy for any service client data use.

Service derived data-use Cloud Computing Explains what derived data is produced by the SP from ser-

vice client data.

Data mirroring/backup

Data mirroring latency Cloud Computing Indicates the difference between the time data is located on

mirrored-storage and the time which the same data is located

on primary-storage.

Data backup method Computer Networking, Cloud

Computing

Indicates a list of methodologies that are used to make a

backup service client’s data.

Frequency of data backup Computer Networking, Cloud

Computing

Explains the time interval in between full backups of service

client’s data.

Backup retention time Cloud Computing Indicates how long a designated backup exists to be utilised

for data restoration.

Backup generations Computer Networking, Cloud

Computing

Explains the total number of backup generations existing to

be utilised for data restoration..

Maximum data restoration-

time

Computer Networking, Cloud

Computing

Explains the committed time that is taken to restore service

customer data from a backup.

Successful data restorations

rate

Computer Networking, Cloud

Computing

Indicates the committed data restorations success rate, ex-

pressed as the total number of error-free data restorations car-

ried out for the clients upon the aggregate number of data

restorations (in percentage).

Data life-cycle

Type of data deletion Cloud Computing Refers to the quality of data elimination, ranging from the

weak elimination that only remove the reference to the data,

to strong sanitisation methods to ensure that removed data

cannot be easily recovered.

Timely effective deletions rate Cloud Computing Describes the total number of the requests for data elimina-

tion by client accomplished during a pre-deﬁned period of

time limit over the whole number of requests for data dele-

tion (in percentage).

Continued on next page

Table 5 – continued from previous page

SLO Category Relevant SLO Service Domain Description

Tested storage retrieve-ability

rate

Cloud Computing Indicates the amount of veriﬁed clients data to be recoverable

within the measurement period, right after the data has been

removed.

Data portability

Data portability format Cloud Computing, SDN, SFC Refers to the e-format(s) that service client data may get ap-

proached by or transferred to the service.

Data portability interface Cloud Computing, SDN, SFC Refers to the methods that can be utilised to transfer cus-

tomers’ data to the service or from the service.

Data transfer rate Cloud Computing, VPN,

SDN, SFC

Indicates the minimum rate of service customer data that can

be transferred from and/or to the service through using the

method(s) offered in the data interface.

Table 6: Personal data protection service level objectives

SLO Category Relevant SLO Service Domain Description

Conduct codes

Applicable data protection

codes of conduct, certiﬁca-

tions, standards

Cloud Computing, VPN,

SDN, SFC

Refers to a list of certiﬁcation mechanisms and methods, stan-

dards and the data protection codes of conduct that the service

agreed with.

Purpose speciﬁcation

Processing purposes Cloud Computing, SDN, SFC Indicates the list of purposes of processing which are beyond

the requests by the clients acting as a controller.

Data minimisation

Period of temporary data re-

tention

Cloud Computing, SDN, SFC Indicates the utmost duration in which provisional data is

maintained after recognition that the provisional data is un-

used.

Period of service customer

data retention

Cloud Computing, SDN, SFC Describes the utmost duration that a service client data is

maintained right before destruction by SP and after acknowl-

edgement of a request to erase the data or contract termina-

tion.

Use limitation

Indicates number of customer

data law enforcement disclo-

sures

Cloud Computing Indicates the number of private data disclosures to law en-

forcement authorities over a predeﬁned period of time.

Personal data disclosure noti-

ﬁcations

Cloud Computing Describes the number of personal and private data disclosures

to law enforcement authorities notiﬁed to the clients over a

certain period of time.

Transparency

Contractors and subcontrac-

tors lists

Cloud Computing Indicates the SP’s contractors and subcontractors that partici-

pate in the processing of the client’s private data service.

Special data categories Cloud Computing Indicates the list of the particular categories of personal data

such as ﬁnancial and health related data or sensitive data.

Accountability

Documentation Cloud Computing, VPN,

SDN, SFC

Describes a list of the documents which is made available

by the SP in order to demonstrate the admission to the data

protection obligations and needs.

Personal data breach policy Cloud Computing, VPN,

SDN, SFC

Refers to the policy of the SP regarding data breach.

Data location

Data geolocation list Cloud Computing Determines the geographical position where the clients’ data

processed and also stored by the SP.

Selection data geolocation Cloud Computing Describes whether clients can choose an speciﬁc geographi-

cal site to store the service client data.

Intervene-ability

Access request response time Cloud Computing, SDN, SFC Describes the period of time in which the SP must communi-

cate the essential information to permit the client to respond

to access-requests via the data subjects.

3 Service Quality Categories & QSD

Although all quality requirement parameters

(QoX) play a signiﬁcant role in SLA, within an

agreed/standard service life-cycle, Fig. 3 (adapted

from [57]) does not show QoX and QoS is the only

one that is taken into account, what is quite insufﬁcient

nowadays. Quality documents are necessary for each

and every one of the activities of the reference service

life-cycle and their importance is explained below:

Figure 3: Service life-cycle, adapted from [57]

• Advertisement: EUs and SPs publish and/or ex-

change the quality requests and offers, respec-

tively. Both quality documents are named as

”Quality-Based Service Descriptions (QSDs)".

• Matchmaking: QSDs are matched up in order to

determine if offers are capable to support the re-

quirements of EUs. The consequence is that the

advertised functionally equivalent services are ﬁl-

tered and selected based on their ability to satisfy

the EU quality requirements afterwards.

• Negotiation: QSDs or SLA templates are inter-

changed between SP and EU as well as between

InP and SP. The feasible agreement between the

actors involved leads to the deﬁnition of an SLA.

• Monitoring/Assessment: the SLA is tracked in or-

der to discover EU and/or SPs’ violations of their

quality and functional terms and conditions.

• Adaptation: The adaptation/recovery and proac-

tive functions can be taken in case the SLA is un-

fulﬁlled. A feasible recovery action might need the

matchmaking activity execution or a further SLA

negotiation to discover an alternative service. It

may also occur that an alert is sent to the evaluation

component of the monitoring activity that pursues

to execute.

QSDs are used in the ﬁrst two and/or three service

life-cycle activities, while SLAs are utilised in the last

three service life-cycle activities. Therefore, there is not

any standard and uniform document for the quality to be

utilised among entire activities mentioned in the service

life-cycle and this is a main drawback which is time

consuming, as transformations of the documents must

take place to another format from an origin template.

Although several different QoS categories and at-

tributes can be found in different research reports, it is

possible to extract few of them that appear most often

and can be considered as essential QoS categories and

attributes, respectively. The other attributes are mostly

context-dependent (i.e., quite particular) or capture sec-

ondary features, as they appear very seldom in the pro-

posed Services Quality Managements (SQMs) so far.

Therefore, the most reported QoS categories and at-

tributes can be considered as the most important ones.

As shown in Table 7, throughput, latency, process-

ing time and response time are the attributes that often

demonstrate the category of performance, that is avail-

able in the majority of SQM proposals. Although it is

not a main focus of this survey, security is the other

signiﬁcant category that has three foremost attributes

among others, namely; authorization, authentication,

and non-repudiation, which are steadily existent in the

SQMs. Reliability, accuracy as well as availability are

the other three most signiﬁcant attributes that are not

categorised under a certain service quality category.

Data quality features are seldom considered in

SQMs. The only data quality feature, that is frequently

considered in reported articles, is ”correctness". The ac-

curacy attribute deﬁned by various contributions refers

to the correctness of the service and also of the data

that it provides. Since the output of a service is often

compound of information, data quality may considered

as a part of the QoS, and it can drive thoroughly the

analysis of the provided output as well as necessary in-

put. Data quality is a multidimensional concept that

deﬁnes the suitability of the utilised data for the spe-

ciﬁc service in which they are involved [57][58]. The

utmost signiﬁcant and representative attributes of data

quality, which should be part of a SQM are accuracy,

completeness, consistency, and timeliness. Among the

reviewed articles, [90] is a quite old reference but it is

the most cited article in research papers along the US

patent (Pub. No.: WO2001099349) for the QoS spec-

iﬁcations area. These attributes can be utilised by the

service for the data correctness investigation purpose,

the existence of inconsistent or missing values, and the

information updateness.

Some of the proposed SQMs take into consideration

speciﬁc network aspects. Usually, there is a network

quality category in these SQMs that comprises the four

most frequent attributes of network, namely: packet

loss ratio, jitter, network delay and bandwidth. Table 7

lists the attributes in service quality categories along the

list of references for published approaches. However,

most of the attributes are addressed in [57] comprehen-

sively.

4 SLA Methodologies

The preparation and deployment of a robust method

that combines both economical and technical aspects in

order to offer a contract to satisfy all actors in an SLA

scenario is an important task to be done. Hereafter,

in this article, a contract that can satisfy EU, SP and

SP requirements is named a win-win contract and

it may not be the purpose of all service providers.

However the analysis and conclusions of this work can

be applied to all contexts. Although some methods

have been reported and published for SLAs, yet a lack

of a comprehensive model makes this research domain

open for further research. In this section we review

available methods proposed so far and we categorise

them in dynamic and static methods and also on

decision-making based methods.

4.1 Dynamic & Static Methodologies

There are signiﬁcant differences between behaviours of

run-time service as well as the expectations of quality

stated in the contract with the clients due to the informal

SLAs’ nature. Managing SLAs in ICT service indus-

tries could be handled through either Static or Dynamic

methods [48][25][12]. Although there are challenges

in developing methods for static SLAs to consider all

QoX parameters rather than choosing the only QoS, the

development of dynamic methods is an even more criti-

cal issue for SPs. This is because of the emerging tech-

nologies as well as the continuous and frequent changes

in service needs and techniques over the time. Any

SLA updating to meet any desired change in a service

requires re-mapping and reformulation of all criteria.

Recently, dynamic solutions are becoming increasingly

commonplace as SLA methods and most of the research

articles focus on recent technologies such as cloud com-

puting. As an instance, W. Halboob et al., published 23

contributed research articles in this context [44][45].

Table 7: Service quality categories and attributes in service quality

categories’ approaches (Adapted from [57])

Category Attributes SQM approaches refer-

ences

Performance

Response time [57][58][73][87][22][2][84]

[66][17][36][69]

Processing time [57][58][73][84][66][17][68]

Latency [57][58][73][87][22][84][17]

[36]

Timeliness [57][22][84][90]

Precision [57][58][90]

Throughput [57][73][87][22][2][84][17]

[36][69]

Security

Authentication [57][58][87][2][84][66][17]

[69]

Authorization [57][58][87][2][66][17][36]

[69]

Security Level [57][58][66][90]

Integrity [57][58][73][87][2][66][17]

Conﬁdentiality [57][58][73][87][2][84][66]

[17][36][69][90]

Accountability [57][58][87][66][17][36][69]

Traceability [57][58][87][2][17][36][69]

Non repudiation [57][58][87][2][17][36][69]

Data encryption [57][58][87][2][84][17][36]

[69]

Isolation [57]

Data

Maturity/age [57][66][17]

Timeliness [57][58][66][17]

Reliability [57][66][17]

Completeness [57][58][66][17]

Conﬁguration

Virtual organisa-

tion

[57][84][66]

Location [57][66]

Level of service [57][17][90]

Service version [57][66]

Supported stan-

dard

[57][58][73][87][22][2][66]

[17][69]

Uncategorised

Cost [57][58][87][84][66][17][36]

[68][90]

Availability [57][58][73][87][22][2][84]

[66][17][36][69]

Accessibility [57][58][73][2][84][66][17]

[36][69]

Accuracy [57][58][73][22][84][66][17]

[36][69][90]

Reliability [57][58][73][87][22][84][66]

[17][36]

Capacity [57][58][73][87][66][17][36]

Believability [57][66]

Maintainability [57][58][22][66]

Relative impor-

tance

[57][90]

Complexity [57][22]

Customer service [57][22]

Dependability [57][58][73][22][84][17]

Stability [57][58][73][87][22][2][17]

[69]

Trust [57][22][68]

Understandability [57][22][17]

Integrability [57]

Interoperability [57][73][84]

Resource efﬁ-

ciency

[57]

Re-usability [57]

Scalability [57][73][87][84][17]

4.2 Decision-Making

The services industry in an ICT domain has been ar-

rived to be a dominating activity in most advanced in-

dustrialized economies. Generally, beyond 75% of the

labour force is engaged in the domain of services indus-

try in the United States alone, with an overall produc-

tion demonstrating about 70% of whole industry out-

put [59][31] and, simultaneously, ICT service indus-

tries/providers are growing signiﬁcantly fast nowadays.

The effective and efﬁcient management of ser-

vice providing processes of decision-making within the

bounds of ICT industry is absolutely important to any

SP in practice. However, an efﬁcient decision-making

is even difﬁcult and very complex to achieve due to the

various quality requirements from EUs point of view.

Particularly this is the case for the processes of complex

decision-making, involved in the ICT services delivery,

extremely uncertain and potentially volatile client re-

quirements, complex models of SLA, and also the ne-

cessity to capture diverse feasible aspects of ICT ser-

vices delivery such as QoX. We can divide decision-

making processes into two: i) General decision-making

and ii) Decision-making under uncertainty categories.

Furthermore, the decision-making under uncertainty

has two subcategories that are deﬁned as: i) Gen-

eral decision-making under uncertainty and ii) Spe-

ciﬁc decision-making under uncertainty, and most of

SLA decision-making among SP, EU and InP could

be utilised and accommodate under these two subcat-

egories.

4.2.1 General Decision-Making

Recently research about SLA and its management has

been soared because of the importance of the SLAs

in ICT industries. A lot of the work deals with au-

tomated and also technically oriented runtime intelli-

gent SLA negotiation [128][99][34]. Although most

of the recent works did not emphasised and concen-

trate on QoX, they highlighted a key challenge in this

area as the prediction of cost and QoS, as to be able

to make an optimal decision already in the negotiation

phase [34][116][26].

Once negotiations are accomplished and the SLA is

deployed, the SP should be able to implement provision

according to different policies, and ﬁnding such policies

so as to achieve a win-win deal is another important area

of research. As an instance, SPs can allocate resources

to guarantee SLA fulﬁlment (guaranteed enforcement),

or exclusively allocate some resources per need (lazy

enforcement) [34][3]. Moreover, the risk and its man-

agement are important phases on SLAs and it signiﬁ-

cantly effects on business due to service downtime. For

example T. Setzer et al. [95], proposed a model for op-

timal service-window scheduling to minimize the busi-

ness impact, but unplanned outages are not considered

in this research and it might not be able to achieve a

win-win deal and negotiation.

Another signiﬁcant research domain in SLA risk-

management is evaluation of the probability that SLOs

speciﬁed in the SLA will not be met as reported in

[105][106]. Moreover, to develop such models, knowl-

edge of the involved statistical distributions is crucial

[35].

4.2.2 General Decision-Making Under Uncertainty

The general decision-making approaches under uncer-

tainty concern various ﬁelds such as cognitive sci-

ence, engineering, artiﬁcial intelligence, economics and

many others. However, it is out of the scope of this

paper to review the vast literature on decision-making

theories under uncertainty, but in general, most of the

proposed solutions are based on either decision-making

under risk [52] or strict-penalty [51]. There are some

reviews about this topic such as [112] and however be-

ing a bit old, it covers most of the fundamental required

knowledge. In this area, a study of how the formulation

of decision-making under uncertainty can induce devia-

tions from maximizing expected monetary value found

that even with small expected values differences, sub-

jects on average did maximize expected value in their

choices [51]. Another study reported that maximiza-

tion of expected value to be the best explanation of sub-

ject behaviour in so called duplex gambles in the losing

form (where the subject can lose but not win money),

but not in the winning form (where the subject can win

but not lose money) [24].

However, it must be noted that the purpose of this

survey is not to contribute to the general descriptive

question of decision-making under risk/strict-penalty

and the purpose of this section is just to introduce SLA

decision-making to be considered to use in the SLA

methodology development.

4.2.3 Speciﬁc/Particular Decision-Making Under

Uncertainty

Beyond the general studies of decision-making under

uncertainty, also a number of more speciﬁc studies have

been conducted to describe particular contexts or par-

ticular stakeholders such as entrepreneurs. One study

addresses the decision-making of CEOs [67]. Their

behaviour seems to deviate a lot from expected util-

ity maximization. Another research, looking at en-

trepreneurs in China, concludes that even though the

entrepreneurs are more willing to accept strategic un-

certainty (related to competition and trust), they do not

differ from the control group when it comes to non-

strategic uncertainty, such as risk [50].

These results are consistent with a study of en-

trepreneurs in Denmark, who do not seem to be more

or less risk-averse than the total population, nonethe-

less they are rather optimistic regarding the chance of

occurrence for the best consequence in lotteries with

real money [5]. In [62], M. Lefebvre et al, report that

the risk taking of executives under different incentive

contracts and the impact of different incentive contracts

(stock options vs. stock shares) for executives was

investigated using students. The results indicate lots

of excessive risk-taking, in particular risk-seeking be-

haviour in the face of small probability gains and large

probability losses.

The comprehensive literature on decision-making

under uncertainty in speciﬁc/particular contexts, some

of which has been cited above, indicates that such con-

textual investigations are considered worthwhile. There

is no simple and/or straightforward ways to generalise

across contexts, instead, a piecemeal approach is obvi-

ous in the literature. It requires to be emphasised that

the ICT SLA decision-making context has, so far, not

been investigated.

5 Modelling and Architecture

A successful SLA requires a comprehensive modelling

and a robust architecture. The SLA model develops a

sincere functional service description to permit for the

expression of service quality guarantees as well as non-

functional service properties. Furthermore, SLA comes

up with the architecture along with an speciﬁcation of

the quality characteristics that the service will provide.

This speciﬁcation allows the architecture to pick out

the service that best supports the system’s quality at-

tribute requirements. This section presents the research

ﬁndings and reports on SLA modelling and architecture

along with the monitoring which is one of the important

objectives for SLA.

5.1 SLA Modelling

Yixin Diao et al., proposed a modelling framework that

uses queueing-model-based approaches for estimation

of the impact of SLAs on the delivery cost. Further-

more, they proposed a set of approximation techniques

to address the complexity of service delivery and an

optimization model to predict the delivery cost subject

to service-level constraints and service stability condi-

tions [26]. Another recent work on this context intro-

duced new SLA scenarios and considered new quality

parameters in SLA modelling. Moreover, the use of

SDN paradigm has been proposed and discussed to im-

plement the fulﬁlment of the SLAs established among

the actors [93].

The SLA negotiation is an essential mechanism to

guarantee the performance of supplied service and to

enhance the trust between EUs and SPs. In this area

there are studies published that focus on the negotia-

tion part of SLA. As an instance, in [39] an outline for

the deﬁnition of a Service Level Speciﬁcations (SLSs)

format, the semantics, and some early ideas on the re-

quirements of negotiation of SLSs have been discussed.

A mutual negotiation protocol applying an alternative

offers model for resource allocation and scheduling in

Grid-Federation as well as a multi-steps SLA negotia-

tion, that includes the cloud SP selection and the ne-

gotiation with the chosen SP are addressed in [53] and

[7] respectively. To develop a general framework for

strategic negotiation of service level values under time

limitations, the latest developments in agent research

are studied in [99]. Another study proposed a method-

ology which aims to appraise the service performance

at early stages of the development process using simu-

lation. The simulation data may be utilised ﬁrst to ne-

gotiate the SLA preliminary performance of the service

between SPs and EUs looking for a given service, and

later to monitor it [92]. In [100] a framework has been

proposed for strategic negotiation of service level val-

ues under time limitations and exemplify the usage of

the framework through extending the ”Bayesian learn-

ing agent" to cope with the limited duration of a ne-

gotiation session. G. C. Silaghi et al., claimed that

opponent learning strategies are worth to be consid-

ered in open competitive computational grids, leading

to the fair satisfaction of actors and an optimal allo-

cation of resources. On the other hand, for intra- and

inter-domain service level negotiation, an extension of

the COPS protocol has been proposed and named as

COPS-SLS [80].

Among the reviewed studies in SLA domain, there

are articles concentrated on SDN based approaches

[93][70][109][111] and few of them contributed with

new concepts in the area [93]. On the other hand,

with the fast growth of some other technologies such

as cloud computing, researchers became interested in

making more investigation on SLA negotiation, devel-

opment and deployment on the cloud. As an instance, in

[34] authors tried to answer a research question whether

enterprise ICT professionals could maximize expected

value when procuring availability SLAs, and they tried

to explore an optimal solution for it, from the business

point of view.

In general, SLAs are important in cloud comput-

ing as they establish agreements between the cloud cus-

tomers and their service suppliers, concerning the stan-

dard of the provided service. Many research articles re-

ported their ﬁndings on SLA domains over cloud com-

puting and we can categorise these articles in eight cat-

egories: i) SLA negotiation and monitoring; ii) Re-

views on QoS aware SLA; iii) SLA decision-making;

iv) Security in SLA; v) Speciﬁc theories based SLA;

vi) Energy-efﬁcient SLA solutions; vii) Cloud based ar-

chitecture and modelling and viii) SLA aware resource

allocation.

In the ﬁrst category, multi-step negotiation of SLA,

that includes the selection of cloud providers and

the negotiation with the selected SP is introduced

in [7]. However, [61] tried to come out with an opti-

mal solution for infrastructure service under the Euro-

pean project called OPTIMIS. Furthermore, negotiating

frameworks for SLA of cloud-based services are pro-

posed by [128][99][111][27]. In fact, SPs hesitate to

negotiate and instead providers prefer to offer strictly

binding SLAs. It is because of the economic risk assess-

ment exposure which is a signiﬁcant challenge. There-

fore it requires some speciﬁc models to bring the risk-

aware SLA solutions such as [49][76][71]. Although

most of the articles considered QoS parameter to val-

idate their ﬁndings, there are signiﬁcantly small num-

bers of articles that rely on QoE instead [121][33][91].

One of the main challenges in the monitor-

ing is monitoring violations. This is because

the existing computer networking platforms allow

users to build distributed, large, and complex ap-

plications. Thus, it is quite critical to develop

SLA monitoring solutions prediction and preven-

tion of SLA violations. Here are some examples:

[116][19][96][64][63][65][30][28][78][126].

In the review process we found few review articles

and an SLA handbook containing SLA research ﬁnd-

ings for cloud environment and they are worth read-

ing [18][77][10]. Moreover, some research focuses

on decision-making process in SLA for cloud comput-

ing such as [34][6]. The paradigm of cloud comput-

ing guarantees trustworthy services, which is available

through any place within the world, in associate degree

on-demand manner. To adopt cloud services, insufﬁ-

cient security has been already known as a signiﬁcant

obstacle. To accommodate the risks related to outsourc-

ing knowledge and applications to the cloud, new ways

for security assurance are much required. Therefore,

some research has been done on SLA security in the

cloud. For instance they propose authentication inter-

face to access a cloud service [13][9][8][15]. In some

published research reports, speciﬁc theories have been

utilised for SLA design and development. In [116], the

authors utilised Bayesian Model to propose an approach

for predicting SLA violations, which uses measured

datasets (QoS of used services) as input for a predic-

tion model. Another research used the advantages of the

Game Theory to model an SLA negotiation and frame-

work for the QoS assurance purpose within the clouds

federation [32][29]. Furthermore, in [97] data-driven

probabilistic has been considered for modelling appli-

cation resource demand for resource allocation and they

claimed that more than 50% savings are demonstrated

using the proposed approach for resource allocation in

the Yahoo’s data centres.

Renewable energy plays a crucial role for assistance

of meeting basic energy needs through the use of mod-

ern technologies known as GreenTech. Therefore, with

the large-scale deployment of either virtualised or phys-

ical data centres, energy consumption and SLA have

become the immediate issue to be solved. That is why

some researchers aimed to ﬁnd energy-efﬁcient SLA

solutions. However, yet there are small number of pub-

lished articles in this area and this research domain

is trending in recent years [109][37][47][4] [41][16].

Generally, in terms of SLA architecture there is not suf-

ﬁcient resources and this constraint goes to the SLA ar-

chitecture for cloud computing as well. In [101], the

authors outlined a value model followed by an SLA

with a partial utility-driven scheduling architecture, that

comes with the partial utility the consumer offers to a

speciﬁc level of depreciation.

SLA-aware resource allocation in cloud environ-

ment has gained more interest to pursue research on

its open challenges and many research articles have

been published in this domain. In [53], as an exam-

ple, a mutual negotiation protocol that is using the al-

ternate offers model has been proposed for resource al-

location and scheduling in Grid Federation and rele-

vant issues and challenges of SLA-aware resource al-

location are discussed in [18]. In [37], the authors pro-

posed a dynamic resource management based on queu-

ing theory through integer programming, control the-

ory techniques and integrating the timing-analysis to

schedule Virtual Machines (VMs) which is driven by

range-based non-linear reductions of utility, different

for classes of clients and among different ranges of

resource allocations [102] and another approach com-

putes the resource allocation ratio based on the histori-

cal monitoring data from the online analysis of the host

and network utilisation but any pre-knowledge of work-

loads [109]. Furthermore, there are existing works pub-

lished with virtualised and proactive approaches such

as [123][78] and also research challenges and pub-

lished solutions so far are discussed in [15]. An SLA

based cloud computing that facilitates resource alloca-

tion which takes into account geographical location and

the workload of distributed data centres is proposed in

[110]. An efﬁcient heuristic algorithm based on dy-

namic programming and convex optimization is intro-

duced to solve the mentioned resource allocation prob-

lem by H. Goudarzi et al. [41] and yet there are several

articles published in this domain for cloud computing

such as [103][120][42][38][117][129], for further read-

ing.

5.2 SLA Architecture

Typical SLAs are only characterised at a single layer

and do not provide insight into parameters or metrics

at the various service stack lower-layers. Therefore,

they do not permit providers to manage their service

stack optimally. In [46], a reference architecture is pro-

posed for a multilevel SLA management framework.

All fundamental concepts of the proposed framework

have been discussed in this study and its main architec-

tural interactions and components explained in detail.

Many challenges of SLA-oriented resource alloca-

tion have been addressed in data centres to satisfy com-

peting applications requests for computing services. In

this study, the authors proposed an SLA-oriented archi-

tecture to overcome resource allocation challenges in

data centres. They claimed that the proposed frame-

work can be effectively implemented using the pro-

posed ”Aneka platform".

It requires to be emphasised that the SLA architec-

ture context has, so far, not been investigated well and

there are not enough articles in this topic. Therefore, we

expect that, it would be one of potential future research

directions in this context.

5.3 SLA Monitoring

One of the SLA objectives is SLA monitoring. An

SLA monitoring must be accomplished right after the

contractual agreement and SLA deployment to meet

the EUs expectations. The service monitoring is sig-

niﬁcantly important to follow the progress of its per-

formance and to ensure that the service complies with

the agreed SLA. The SLA monitoring is accomplished

by applying several statistics like monitoring data,

analysing, a systematic process of collecting and other

factors that derive the higher value from the business.

Once done with the monitoring, it proceeds with the re-

porting of SLA where the SP is capable to clearly see

the dashboard breakdowns with time, policy and status

where it can recognise the areas of the problem. Both

SLA monitoring and reporting always assist to meet

the agreement for business applications and provide the

highest possible performance.

In [19], it is shown that the dynamic, diverse and

unforeseeable nature of both application workloads

and cloud services make quality-assured provision of

such cloud service-based applications (CSBAs) a main

challenge. The authors propose a cross-layer frame-

work for SLA monitoring and its main aspects con-

tain: (a) realtime, ﬁne-grained visibility into CSBA per-

formance, (b) visual descriptive analytics to identify

correlations and interdependencies between cross-layer

performance metrics, (c) temporal proﬁling of CSBA

performance, (d) proactive monitoring, detection and

root-cause analysis of SLA violation, and (e) support

for both reactive and proactive adaptation in support

of quality-assured CSBA provision. The proposed ap-

proach for SLA monitoring is validated by a prototype

implementation.

Joel Sommers et al., in [108], active measure-

ments for the SLA performance metrics are uniﬁed in

a discrete time-based tool called SLAM (SLA Moni-

tor). The SLAM tool implemented to carry out multi-

objective probing with two different topologies. SLAM

transmits UDP packets in a one-way manner between

a transmitter and receiver. SLA monitoring is of sig-

niﬁcant interest to both EUs and SPs to ensure that the

network is operating within the acceptable bounds. As

discussed in the article, the obtained results illustrate

that standard techniques for measuring loss rate, delay

variation and end-to-end delay cannot provide an accu-

rate approximation of the state of the network, thereby

preventing an accurate assessment of SLA compliance.

Among event based SLA monitoring, [64] proposed a

monitoring approach called prevention and prediction

based on an event monitoring (PREVENT) system, a

framework for prediction of run-time and further pre-

vention of violations. PREVENT is based on the idea

of monitoring as well as analysing run-time data to trig-

ger adaptation actions in endangered composition in-

stances.

The ”Web Service Level Agreement (WSLA)"

framework is targeted at describing and monitoring

SLAs for Web services. Although, WSLA has been de-

veloped for an environment of Web services, it is also

enforceable to any other inter-domain management sce-

nario, like service management and business process, or

the management of systems, applications and networks

in general. Upon the receipt of an speciﬁcation of SLA,

automatically the WSLA monitoring services are con-

ﬁgured to enforce the SLA. An implementation of the

WSLA framework termed SLA Compliance Monitor,

is publicly available as part of the IBM Web Services

Tool-kit [54].

The SLA monitoring is playing a key role to ensure

compliance with the stated terms of the contract. In

order to make an appropriate system for service levels

measurement, it is very important to realise what is re-

alistically feasible as far as service is concerned. The

technical metrics do have a role to play in consideration

of service levels. The needs to be considered in this

stage soonest response time, scalability, delay and jit-

ter for QoS and other aspects required to be taken into

account for QoX in future.

6 Open Problems and Future Research Direc-

tion

To make reading the SLA-related open problems easy,

we proposed ﬁve categories to cover existing chal-

lenges: a) Negotiation procedures, b) Methodology and

modelling, c) Implementation and deployment, d) SLA

monitoring and e) Security.

6.1 The SLA Negotiation Procedures

Five speciﬁc open challenges are identiﬁed for the SLA

negotiation protocols and processes. First and fore-

most, SLA negotiation processes and protocols diver-

sity constrain the negotiation for establishing new SLAs

as it was initially proposed by [125][124][127]. More-

over, as the SLA is the pre-set agreement among the

actors, the implemented SLA modiﬁcation, and SLA

negotiation between separated administrative domains

are not an easy process and most likely it is impossi-

ble [125][124].

Often SLAs are technical documents regarding to

terminology and concepts which may only be under-

stood by a minor class of technology oriented special-

ists. Therefore evaluation and improvement do not take

place on a regular basis. Such "dead-end” documents

have a very restricted meaning for EUs and their man-

agement as it was initially proposed by [118]. On the

other hand, unclear service speciﬁcations can be con-

sidered as an open challenge and yet there is a need of

protocol to develop a comprehensive SLA. For instance,

agreements on "the availability of a network” are gen-

erally determined using a metric called the percentage

of availability. It is extremely tough to specify what the

accurate meaning of such a metric is in the context of a

speciﬁc business location [118]. Although some recent

articles proposed frameworks considering SLA with ac-

curacy such as [86], yet further research is required to

be done. By providing such accuracy in the agreement,

could the provider answer what is the difference and/or

differences between an availability percentage of 98%

and 99%? The agreement can also deﬁne well whether

this 98% is on a yearly basis or not. If it is not clearly

deﬁned in the SLA, it means that the network is allowed

to go "Down” for a whole week, after being "Up” for

the last 51 weeks!

6.2 Methodology & Modelling

SLA is a bilateral agreement between two actors and

each involved party must have a level of satisfaction

achieved in this negotiation (the win-win scenario).

The admission control policy is another open problem

in SLA methodology and modelling development that

needs more investigation and research because, from

the SP point of view, a decision on which client de-

mand to accept affects the reputation of the SP, proﬁt

and performance [125][124].

Most of proposed methods developed are based on

either arbitrary assumptions and parameters or very

general parameters such as 99.99% availability percent-

age! These would be ﬁne for overall evaluation and/or

documentation purpose, but the problem reveals when

it goes for a real-time implementation. Therefore, deﬁ-

nition of more realistic SLAs should be researched be-

cause it may lead to a better use of the network assets

and resources.

With the variety of services provided by SPs, be-

sides the QoS, other quality requirements, named QoX,

such as Quality of Experience (QoE), Quality of Infor-

mation (QoI), discussed in Section 3, must be taken

into the account in new SLA modelling and establish-

ment [93].

6.3 Implementation and Deployment

There are both practical and theoretical constraints to

SD-SLA (Software-Deﬁned SLA) [93][70][111][109].

Beyond the cost and physical limitations that are al-

ways system-level parameters that need to be managed,

poorly designed SD-SLAs may not be able to be imple-

mented on software-deﬁned networks to support pro-

vided services such as cloud services. As an instance, if

essential operations are serialised, afterwards, they can-

not be programmatically scaled out and up to satisfy

an SD-SLA. By developing and implementing appro-

priate SD-SLAs, there are chances to step into a contin-

uous model for many signiﬁcant background-processes,

that are previously required to be scheduled due to the

limitations of ﬁxed resources [60]. Furthermore, yet

there are further opportunities to develop a program-

matic SD-SLA validation through automated test ana-

lytic and infrastructure [60][14].

The performance forecast management enables the

recommendation for performance improvement and op-

timisation, therefore it can be considered as an open

question in utility computing environments. Moreover,

dynamic management of resource allocation has to be

considered in the implementation of SLAs, because it

addresses which resource is the best and appropriate

for a current admitted request from the point of view

of both actors [125][124].

6.4 SLA Monitoring

SLA monitoring measures an SLA compliance and

monitors true uptime for provided services which is es-

sential for better service delivery and network monitor-

ing is utilised to ensure the hosts and nodes connec-

tions with the speciﬁed bandwidth as well as monitor-

ing the proper packet delivery. The SLA should be es-

tablished between providers and end-users from a dif-

ferent end-to-end point of view. As an instance, if the

service of system has been outsourced, not only from

SP to a EU but also from one SP to another SP, there

must be an SLA agreement between them and this re-

quires a dynamic SLA monitoring solution as well to

help in managing the network and to maintain the SLA

requirements [74][125][107].

6.5 Security

Security capabilities measurement which can be guar-

anteed and quantiﬁed for SLA contexts is one of im-

portant challenges in expressing security properties.

Capability of security is a combination of mutually-

reinforcing security controls to mitigate risks as well as

demonstrate compliance with the EUs security require-

ments [83][82][89].

The lack of trust established in SLA agreement

about particular security capabilities offered and guar-

anteed by the SPs is an important challenge faced

by SLA actors. Several certiﬁcations have been pro-

posed in cloud-computing and outsourcing environ-

ments. Nonetheless, these certiﬁcation schemes are

largely inappropriate in the service provisioning context

and they do not ensure better security [83][82][85].

In the context of SLAs, metrics can be deﬁned

and utilised to measure and track the compliance

of security-related SLAs. However, security-related

SLAs’ metrics are not well-established so far, and avail-

able SLA metrics are typically measured and deﬁned

based on implemented and statistics with appropriate

provisions and capabilities related to EUs and SPs.

Therefore, effective security metrics development for

SLAs has proven to be very challenging [83][94][115].

Table 8 highlights and classiﬁes the identiﬁed open

problems in SLA research domain and ﬁve categories

are addressed in the table as SLA Negotiation, Method-

ology & Modelling, Implementation & Deployment,

SLA Management & Monitoring and Security.

Table 8: SLA open problems

Cat. Open Problems References

SLA Negotiation

- The negotiation constrains for es-

tablishing SLA

[125][124][127]

- An implemented SLA modiﬁca-

tion

[125][124]

- SLA negotiation among different

administrative domains

[125][124]

- ”Dead-end” SLA documents [118]

- Unclear service speciﬁcations [118]

Methodology

Modelling

- Lack of other quality requirements

(QoX) consideration

[93]

- Lack of a win-win deal commit-

ment for all network actors

- Admission control policies [125][124]

- Deﬁnition of more realistic SLAs [125][124]

Implementation

Deployment

- Implementation in software-

deﬁned network architecture

[93][70]

[111][109][60]

- Performance forecast manage-

ment in utility computing environ-

ments

[60][14]

- The efﬁcient resource allocation

management

[125][124]

Management

Monitoring

- SLA has to be established among

the providers and clients from vari-

ous end-to-end point of view.

[74][125][124]

Security

- Quantifying security properties in

SLA contexts

[83][82][89]

- Specifying capabilities of security

in SLA contexts

[83][85]

- Evaluating security capabilities

speciﬁed in SLA contexts

[83][94][115]

7 Conclusions

The increasing use of ICT services makes the estab-

lishment of efﬁcient SLA vital for InP, SP and EU to

maintain ongoing contracts. In this survey a large num-

ber of research studies, models and methods were de-

scribed with the aim of revealing the importance of re-

search activities in the SLA domain and it shows that

there is a large research effort to propose fair SLA meth-

ods. Based on the reviewed articles the current SLA re-

search state of art has been identiﬁed in the domain and

a comprehensive SLA conceptual-map was developed

to identify SLAs related concepts. The SLA elements

and actors were introduced and the importance of their

key roles on achieving a successful SLA have been dis-

cussed followed by SLA characteristics and SLO. This

survey was organised in a way to provide a better under-

standing of existing methods, models and architectures

for readers with different research interests related to

SLA and a reading-map is provided in Section 1. Fur-

thermore, SLA relevant application domains are iden-

tiﬁed and discussed to show their applicable scopes in

ICT industries.

Although the proposed methods and architectures

claimed that they were successful in satisfying an ac-

tor with a focus on particular service, yet the lack

of a generic SLA method requires further research to

achieve win-win SLA deployment among all involved

actors. This survey pointed out existing research gaps in

utilising SLA elements to develop a generic methodol-

ogy for SLAs establishment, deployment, management

and particularly that will lead to a win-win situation for

all involved actors.

Finally, various open research problems and future

research directions were discussed and classiﬁed in ﬁve

categories in order to encourage researchers in further

research on developing generic SLA models and archi-

tectures. This study is part of an ongoing SLA project

and is an active research on how to specify and develop

SLAs to achieve win-win agreements among all actors.

A SLA Conceptual-Map

A concept map or conceptual-map could be a diagram

that depicts prompt relationships between ideas. It is a

graphical tool that is utilised to organise and to struc-

ture knowledge. A conceptual map usually represents

concepts and information. The connection between

ideas are often articulated by means of predicates. The

conceptual maps are developed to guide speciﬁc re-

searches. Therefore, for a same topic, SLA for example,

different conceptual maps may be developed to achieve

different research objectives.

Figure 4 illustrates the proposed conceptual-map to

achieve a comprehensive and a generic SLA methodol-

ogy for computer and communication networks, and to

take all quality requirements into account in SLA de-

ployment. The map starts in the top centre box iden-

tiﬁed as Data Communication Networks and it is com-

posed by network entities that can get beneﬁts of SLA

to provide promised level of service for EU and/or SP.

To achieve a desirable SLA, the model must be de-

ﬁned based on SLOs and contains QSDs. In this pro-

posed conceptual-map, all relevant concepts such as

SLA monitoring and requirements engineering are con-

sidered to provide necessary feedbacks and to supply

required information as entries for an SLA methodol-

ogy development.

In the proposed map, all relevant concepts are or-

ganised in a way to achieve A generic SLA methodol-

ogy (shown in yellow box in the middle of the map)

that can provide a win-win agreement among all actors

by considering their QoX requirements. A generic SLA

methodology requires suitable resource management

and decision-making algorithms. Based on a generic

SLA methodology, an SLA model and architecture can

be proposed. Eventually, a model and architecture of

SLA can be evaluated and validated before getting de-

ployed.

Figure 4: SLA conceptual-map created using CMap [81]

References

[1] SLA management handbook. volume 4: En-

terprise Perspective. The Open Group, Thames

Tower, 37-45 Station Road, Reading, Berkshire

RG1 1LX, United Kingdom, October 2004.

[2] OASIS therapeutic community (TC) recom-

mended for continued accreditation. Technical

report, 2008.

[3] Aib, I. and Boutaba, R. On leveraging policy-

based management for maximizing business

proﬁt. IEEE Transactions on Network and Ser-

vice Management, 4(3):25–39, dec 2007.

[4] Amokrane, A., Langar, R., Zhani, M. F.,

Boutaba, R., and Pujolle, G. Greenslater: On sat-

isfying green SLAs in distributed clouds. IEEE

Transactions on Network and Service Manage-

ment, 12(3):363–376, sep 2015.

[5] Andersen, S., Girolamo, A. D., Harrison, G. W.,

and Lau, M. I. Risk and time preferences of en-

trepreneurs: evidence from a danish ﬁeld experi-

ment. Theory and Decision, 77(3):341–357, may

2014.

[6] Andrzejak, A., Kondo, D., and Yi, S. De-

cision model for cloud computing under SLA

constraints. In 2010 IEEE International Sym-

posium on Modeling, Analysis and Simulation

of Computer and Telecommunication Systems.

IEEE, aug 2010.

[7] Anithakumari, S. and Chandrasekaran, K. Ne-

gotiation and monitoring of service level agree-

ments in cloud computing services. In Proceed-

ings of the International Conference on Data

Engineering and Communication Technology,

pages 651–659. Springer Singapore, aug 2016.

[8] Bajpai, D., Vardhan, M., Gupta, S., Kumar, R.,

and Kushwaha, D. S. Security service level

agreements based authentication and authoriza-

tion model for accessing cloud services. In Ad-

vances in Computing and Information Technol-

ogy, pages 719–728. Springer Berlin Heidelberg,

2012.

[9] Bajpai, D., Vardhan, M., and Kushwaha, D. S.

Authentication and authorization interface using

security service level agreements for accessing

cloud services. In Communications in Com-

puter and Information Science, pages 370–382.

Springer Berlin Heidelberg, 2012.

[10] Baset, S. A. Cloud SLAs. ACM SIGOPS Oper-

ating Systems Review, 46(2):57, jul 2012.

[11] Baset, S. A. Cloud service level agreement. In

Encyclopedia of Cloud Computing, pages 433–

445. John Wiley & Sons, Ltd, may 2016.

[12] Benlarbi, S. Estimating SLAs availabil-

ity/reliability in multi-services IP networks.

In Service Availability, pages 30–42. Springer

Berlin Heidelberg, 2006.

[13] Bernsmed, K., Jaatun, M. G., and Undheim, A.

Security in service level agreements for cloud

computing. In Proceedings of the 1st Interna-

tional Conference on Cloud Computing and Ser-

vices Science, pages 636–642. SciTePress - Sci-

ence and Technology Publications, 2011.

[14] Bouchenak, S., Chockler, G., Chockler, H., Ghe-

orghe, G., Santos, N., and Shraer, A. Verifying

cloud services. ACM SIGOPS Operating Systems

Review, 47(2):6, jul 2013.

[15] Buyya, R., Garg, S. K., and Calheiros, R. N.

SLA-oriented resource provisioning for cloud

computing: Challenges, architecture, and so-

lutions. In 2011 International Conference on

Cloud and Service Computing. IEEE, dec 2011.

[16] Cao, Z. and Dong, S. Dynamic VM con-

solidation for energy-aware and SLA violation

reduction in cloud computing. In 2012 13th

International Conference on Parallel and Dis-

tributed Computing, Applications and Technolo-

gies. IEEE, dec 2012.

[17] Cappiello, C., Kritikos, K., Metzger, A., Parkin,

M., Pernici, B., Plebani, P., and Treiber, M. A

quality model for service monitoring and adap-

tation. pages 29–42, 01 2009.

[18] Chana, I. and Singh, S. Quality of service and

service level agreements for cloud environments:

Issues and challenges. In Computer Communica-

tions and Networks, pages 51–72. Springer Inter-

national Publishing, 2014.

[19] Chhetri, M. B., Vo, Q. B., and Kowalczyk, R.

CL-SLAM: Cross-layer SLA monitoring frame-

work for cloud service-based applications. In

Proceedings of the 9th International Conference

on Utility and Cloud Computing - UCC '16.

ACM Press, 2016.

[20] Chiba, M., Clemm, A., Medley, S., Salowey, J.,

Thombare, S., and Yedavalli, E. Cisco service-

level assurance protocol. Technical report, jan

2013.

[21] Choudhury, V. and Sabherwal, R. Portfo-

lios of control in outsourced software develop-

ment projects. Information Systems Research,

14(3):291–314, sep 2003.

[22] Colombo, M., Nitto, E. D., Penta, M. D., Dis-

tante, D., and Zuccalà, M. Speaking a com-

mon language: A conceptual model for de-

scribing service-oriented systems. In Service-

Oriented Computing – ICSOC 2007, pages 48–

60. Springer Berlin Heidelberg, 2005.

[23] D’Antonio, S., D’Arienzo, M., Esposito, M.,

Romano, S., and Ventre, G. Managing ser-

vice level agreements in premium IP networks:

a business-oriented approach. Computer Net-

works, 46(6):853–866, dec 2004.

[24] Davenport, W. and Middleton, M. Expectation

theories of decision making for duplex gambles.

Acta Psychologica, 37(3):155–172, jun 1973.

[25] Dey, P., Kundu, A., Naskar, M. K., Mukherjee,

A., and Nasipuri, M. Dynamic multipath band-

width provisioning with jitter, throughput, SLA

constraints in MPLS over WDM network. In

Distributed Computing and Networking, pages

376–391. Springer Berlin Heidelberg, 2010.

[26] Diao, Y., Lam, L., Shwartz, L., and Northcutt,

D. Modeling the impact of service level agree-

ments during service engagement. IEEE Trans-

actions on Network and Service Management,

11(4):431–440, dec 2014.

[27] El-Awadi, R. and Abu-Rizka, M. A frame-

work for negotiating service level agreement of

cloud-based services. Procedia Computer Sci-

ence, 65:940–949, 2015.

[28] Falasi, A. A., Serhani, M. A., and Dssouli, R. A

model for multi-levels SLA monitoring in fed-

erated cloud environment. In 2013 IEEE 10th

International Conference on Ubiquitous Intelli-

gence and Computing and 2013 IEEE 10th Inter-

national Conference on Autonomic and Trusted

Computing. IEEE, dec 2013.

[29] Falasi, A. A., Serhani, M. A., and Hamdouch,

Y. A Game Theory Based Automated SLA Ne-

gotiation Model for Conﬁned Federated Clouds,

pages 102–113. Springer International Publish-

ing, 2016.

[30] Faniyi, F., Bahsoon, R., and Theodoropoulos, G.

A dynamic data-driven simulation approach for

preventing service level agreement violations in

cloud federation. Procedia Computer Science,

9:1167–1176, 2012.

[31] Faulin, J., Juan, A., Grasman, S., and Fry, M. De-

cision Making in Service Industries. CRC Press,

jul 2012.

[32] Figueroa, C., Figueroa, N., Jofre, A., Sahai, A.,

Chen, Y., and Iyer, S. A game theoretic frame-

work for SLA negotiation. HP Laboratories Palo

Alto, Tech. Rep. HPL-2008-5, 2008.

[33] Frangoudis, P. A., Sgora, A., Varela, M., and Ru-

bino, G. Quality-driven optimal SLA selection

for enterprise cloud communications. In 2014

IEEE International Conference on Communica-

tions Workshops (ICC). IEEE, jun 2014.

[34] Franke, U. and Buschle, M. Experimental ev-

idence on decision-making in availability ser-

vice level agreements. IEEE Transactions on

Network and Service Management, 13(1):58–70,

mar 2016.

[35] Franke, U., Holm, H., and Konig, J. The distribu-

tion of time to recovery of enterprise IT services.

IEEE Transactions on Reliability, 63(4):858–

867, dec 2014.

[36] Frutos, H. M., Kotsiopoulos, I., Gonzalez, L.

M. V., and Merino, L. R. Enhancing service

selection by semantic QoS. In Lecture Notes

in Computer Science, pages 565–577. Springer

Berlin Heidelberg, 2009.

[37] Gao, Y., Guan, H., Qi, Z., Song, T., Huan,

F., and Liu, L. Service level agreement based

energy-efﬁcient resource management in cloud

data centers. Computers & Electrical Engineer-

ing, 40(5):1621–1633, jul 2014.

[38] Garg, S. K., Toosi, A. N., Gopalaiyengar, S. K.,

and Buyya, R. SLA-based virtual machine man-

agement for heterogeneous workloads in a cloud

datacenter. Journal of Network and Computer

Applications, 45:108–120, oct 2014.

[39] Goderis, D., Van Den Bosch, S., T’joens,

Y., Poupel, O., Jacquenet, C., Memenios, G.,

Pavlou, G., Egan, R., Grifﬁn, D., Georgatsos, P.,

et al. Service level speciﬁcation semantics, pa-

rameters and negotiation requirements. Techni-

cal report, February 2002.

[40] Goo, Kishore, Rao, and Nam. The role of service

level agreements in relational management of in-

formation technology outsourcing: An empirical

study. MIS Quarterly, 33(1):119, 2009.

[41] Goudarzi, H., Ghasemazar, M., and Pedram, M.

SLA-based optimization of power and migra-

tion cost in cloud computing. In 2012 12th

IEEE/ACM International Symposium on Cluster,

Cloud and Grid Computing (ccgrid 2012). IEEE,

may 2012.

[42] Goudarzi, H. and Pedram, M. Multi-dimensional

SLA-based resource allocation for multi-tier

cloud computing systems. In 2011 IEEE 4th

International Conference on Cloud Computing.

IEEE, jul 2011.

[43] Gozdecki, J., Jajszczyk, A., and Stankiewicz, R.

Quality of service terminology in IP networks.

IEEE Communications Magazine, 41(3):153–

159, mar 2003.

[44] Halboob, W., Abbas, H., Haouam, K., and

Yaseen, A. Dynamically changing service level

agreements (SLAs) management in cloud com-

puting. In Intelligent Computing Methodologies,

pages 434–443. Springer International Publish-

ing, 2014.

[45] Halboob, W., Abbas, H., Khan, M. K., Khan,

F. A., and Pasha, M. A framework to address in-

constant user requirements in cloud SLAs man-

agement. Cluster Computing, 18(1):123–133,

nov 2014.

[46] Happe, J., Theilmann, W., Edmonds, A., and

Kearney, K. T. A reference architecture for

multi-level SLA management. In Service Level

Agreements for Cloud Computing, pages 13–26.

Springer New York, 2011.

[47] Hasan, M. S., Kouki, Y., Ledoux, T., and Pazat,

J.-L. Exploiting renewable sources: When green

SLA becomes a possible reality in cloud com-

puting. IEEE Transactions on Cloud Computing,

5(2):249–262, apr 2017.

[48] He, R., Lin, B., and Li, L. Dynamic service-

level-agreement aware shared-path protection in

WDM mesh networks. Journal of Network

and Computer Applications, 30(2):429–444, apr

2007.

[49] Hedwig, M., Malkowski, S., and Neumann, D.

Risk-aware service level agreement design for

enterprise information systems. In 2012 45th

Hawaii International Conference on System Sci-

ences. IEEE, jan 2012.

[50] Holm, H. J., Opper, S., and Nee, V. En-

trepreneurs under uncertainty: An economic

experiment in china. Management Science,

59(7):1671–1687, jul 2013.

[51] Huck, S. and WeizsÃ

Ccker, G. Risk, complex-

ity, and deviations from expected-value maxi-

mization: Results of a lottery choice experiment.

Journal of Economic Psychology, 20(6):699–

715, dec 1999.

[52] Kahneman, D. and Tversky, A. Prospect theory.

an analysis of decision making under risk. Tech-

nical report, apr 1977.

[53] Kalati, H. and Deldari, H. A bilateral negotiation

mechanism for SLA-based job superscheduling

in grid-federation. In ICCKE 2013. IEEE, oct

2013.

[54] Keller, A. and Ludwig, H. The WSLA frame-

work: Specifying and monitoring service level

agreements for web services. Journal of Network

and Systems Management, 11(1):57–81, 2003.

[55] Kirsch, L. S. Portfolios of control modes and IS

project management. Information Systems Re-

search, 8(3):215–239, sep 1997.

[56] Koh, C., Ang, S., and Straub, D. W. IT outsourc-

ing success: A psychological contract perspec-

tive. Information Systems Research, 15(4):356–

373, dec 2004.

[57] Kritikos, K., Carro, M., Pernici, B., Plebani,

P., Cappiello, C., Comuzzi, M., Benrernou, S.,

Brandic, I., Kertész, A., and Parkin, M. A survey

on service quality description. ACM Computing

Surveys, 46(1):1–58, oct 2013.

[58] Kritikos, K. and Plexousakis, D. Requirements

for QoS-based web service description and dis-

covery. IEEE Transactions on Services Comput-

ing, 2(4):320–337, oct 2009.

[59] Lacey, T. A., Toossi, M., Dubina, K., and

Gensler, A. Projections overview and highlights,

2016–26. Monthly Labor Review, oct 2017.

[60] Lango, J. Toward software-deﬁned SLAs. Com-

munications of the ACM, 57(1):54–60, jan 2014.

[61] Lawrence, A., Djemame, K., Wäldrich, O.,

Ziegler, W., and Zsigri, C. Using service level

agreements for optimising cloud infrastructure

services. In Towards a Service-Based Inter-

net. ServiceWave 2010 Workshops, pages 38–49.

Springer Berlin Heidelberg, 2011.

[62] Lefebvre, M. and Vieider, F. M. Risk taking

of executives under different incentive contracts:

Experimental evidence. Journal of Economic Be-

havior & Organization, 97:27–36, jan 2014.

[63] Leitner, P., Ferner, J., Hummer, W., and Dustdar,

S. Data-driven and automated prediction of ser-

vice level agreement violations in service com-

positions. Distributed and Parallel Databases,

31(3):447–470, apr 2013.

[64] Leitner, P., Michlmayr, A., Rosenberg, F., and

Dustdar, S. Monitoring, prediction and preven-

tion of SLA violations in composite services.

In 2010 IEEE International Conference on Web

Services. IEEE, jul 2010.

[65] Leitner, P., Michlmayr, A., Rosenberg, F., and

Dustdar, S. Monitoring, prediction and preven-

tion of SLA violations in composite services.

In 2010 IEEE International Conference on Web

Services. IEEE, jul 2010.

[66] linh Truong, H., Samborski, R., and Fahringer, T.

Towards a framework for monitoring and analyz-

ing QoS metrics of grid services. In 2006 Second

IEEE International Conference on e-Science and

Grid Computing (e-Science'06). IEEE, dec 2006.

[67] List, J. A. and Mason, C. F. Are CEOs expected

utility maximizers? Journal of Econometrics,

162(1):114–123, may 2011.

[68] Liu, Y., Ngu, A. H., and Zeng, L. Z. QoS com-

putation and policing in dynamic web service se-

lection. In Proceedings of the 13th international

World Wide Web conference on Alternate track

papers & posters - WWW Alt. '04. ACM Press,

2004.

[69] Mabrouk, N. B., Georgantas, N., and Issarny, V.

A semantic end-to-end QoS model for dynamic

service oriented environments. In 2009 ICSE

Workshop on Principles of Engineering Service

Oriented Systems. IEEE, may 2009.

[70] Machado, C. C., Wickboldt, J. A., Granville,

L. Z., and Schaeffer-Filho, A. ARKHAM: An

advanced reﬁnement toolkit for handling service

level agreements in software-deﬁned network-

ing. Journal of Network and Computer Appli-

cations, 90:1–16, jul 2017.

[71] Macías, M. and Guitart, J. A Risk-Based Model

for Service Level Agreement Differentiation in

Cloud Market Providers, pages 1–15. Springer

Berlin Heidelberg, 2014.

[72] Maheshwari, D. Elements of service level agree-

ment and SLA best practices. Technical report,

Essel Business Excellence Services Ltd., Upper

Ground Floor, Plot No. 19-20, Sector 16-A, Film

City Noida-201301, India, June 2017.

[73] Mani, A. and Nagarajan, A. Understanding qual-

ity of service for web services: Improving the

performance of your web services. 2002.

[74] Marilly, E., Martinot, O., Betge-Brezetz, S., and

Delegue, G. Requirements for service level

agreement management. In IEEE Workshop on

IP Operations and Management. IEEE, 2002.

[75] Martin, J. and Nilsson, A. On service level agree-

ments for IP networks. In Proceedings.Twenty-

First Annual Joint Conference of the IEEE Com-

puter and Communications Societies. IEEE.

[76] Mastroeni, L. and Naldi, M. Compensation

policies and risk in service level agreements:

A value-at-risk approach under the ON-OFF

service model. In Economics of Converged,

Internet-Based Networks, pages 2–13. Springer

Berlin Heidelberg, 2011.

[77] Mirobi, G. J. and Arockiam, L. Service level

agreement in cloud computing: An overview.

In 2015 International Conference on Control,

Instrumentation, Communication and Computa-

tional Technologies (ICCICCT). IEEE, dec 2015.

[78] Morshedlou, H. and Meybodi, M. R. Decreasing

impact of SLA violations:a proactive resource al-

location approachfor cloud computing environ-

ments. IEEE Transactions on Cloud Computing,

2(2):156–167, apr 2014.

[79] Nafarieh, A., Sivakumar, S., Robertson, W., and

Phillips, W. Enhanced adaptive SLA-aware al-

gorithms for provisioning shared mesh optical

networks. Procedia Computer Science, 19:494–

502, 2013.

[80] Nguyen, T. M. T., Boukhatem, N., Doudane, Y.,

and Pujolle, G. COPS-SLS: a service level nego-

tiation protocol for the internet. IEEE Communi-

cations Magazine, 40(5):158–165, may 2002.

[81] Novak, J. D. and Cañas, A. J. The theory un-

derlying concept maps and how to construct and

use them, technical report ihmc cmaptools 2006-

01 rev 2008-01. Technical report, Institute for

Human and Machine Cognition, Pensacola Fl,

32502, January 2008.

[82] Nugraha, Y. and Martin, A. Towards the clas-

siﬁcation of conﬁdentiality capabilities in trust-

worthy service level agreements. In 2017 IEEE

International Conference on Cloud Engineering

(IC2E). IEEE, Apr 2017.

[83] Nugraha, Y. and Martin, A. Understanding trust-

worthy service level agreements: Open problems

and existing solutions. In Camenisch, J. and

Kesdo

gan, D., editors, International Workshop

on Open Problems in Network Security (iNet-

Sec), volume IFIP eCollection-1 of Open Prob-

lems in Network Security, pages 54–70, Rome,

Italy, May 2017.

[84] Pernici, B. Mobile information systems-

infrastructure and design for adaptivity and ﬂex-

ibility (the mais approach). Recherche, 67:02, 01

2006.

[85] Phan, R. C.-W. Review of security engineering:

A guide to building dependable distributed sys-

tems, 2nd edition by ross j. anderson. Cryptolo-

gia, 33(1):102–103, jan 2009.

[86] Rahim, M. A., Haq, I. U., Durad, H., and

Schikuta, E. Generalized SLA enforce-

ment framework using feedback control sys-

tem. In 2015 12th International Conference

on High-capacity Optical Networks and En-

abling/Emerging Technologies (HONET). IEEE,

dec 2015.

[87] Ran, S. A model for web services discovery with

QoS. ACM SIGecom Exchanges, 4(1):1–10, mar

2003.

[88] Richters, J. and Dvorak, C. A framework for

deﬁning the quality of communications services.

IEEE Communications Magazine, 26(10):17–23,

oct 1988.

[89] Ross, R. S. Assessing security and privacy con-

trols in federal information systems and organi-

zations. Technical report, dec 2014.

[90] Sabata, B., Chatterjee, S., Davis, M., Sydir, J.,

and Lawrence, T. Taxonomy for QoS speciﬁca-

tions. In Proceedings Third International Work-

shop on Object-Oriented Real-Time Dependable

Systems. IEEE Comput. Soc.

[91] Sackl, A., Zwickl, P., and Reichl, P. The trouble

with choice: An empirical study to investigate

the inﬂuence of charging strategies and content

selection on QoE. In Proceedings of the 9th In-

ternational Conference on Network and Service

Management (CNSM 2013). IEEE, oct 2013.

[92] Sagbo, K. A. R., Houngue, Y. P. E., and Dami-

ani, E. SLA negotiation and monitoring from

simulation data. In 2016 12th International Con-

ference on Signal-Image Technology & Internet-

Based Systems (SITIS). IEEE, 2016.

[93] Santos-Boada, G., de Almeida Amazonas, J. R.,

and Solé-Pareta, J. Quality of network eco-

nomics optimisation using service level agree-

ment modelling. Transactions on Emerging

Telecommunications Technologies, 27(5):731–

744, feb 2016.

[94] Schneier, B. Secrets and Lies. Wiley Publishing,

Inc., oct 2015.

[95] Setzer, T., Bhattacharya, K., and Ludwig, H.

Decision support for service transition manage-

ment enforce change scheduling by performing

change risk and business impact analysis. In

NOMS 2008 - 2008 IEEE Network Operations

and Management Symposium. IEEE, 2008.

[96] Shenoy, S., Gorinevsky, D., and Laptev, N. Prob-

abilistic modeling of computing demand for ser-

vice level agreement. IEEE Transactions on Ser-

vices Computing, pages 1–1, 2016.

[97] Shenoy, S., Gorinevsky, D., and Laptev, N. Prob-

abilistic modeling of computing demand for ser-

vice level agreement. IEEE Transactions on Ser-

vices Computing, pages 1–1, 2016.

[98] Shetty, N., Schwartz, G., and Walrand, J. Internet

QoS and regulations. IEEE/ACM Transactions

on Networking, 18(6):1725–1737, dec 2010.

[99] Silaghi, G. C., ¸Serban, L. D., and Litan, C. M.

A framework for building intelligent SLA nego-

tiation strategies under time constraints. In Eco-

nomics of Grids, Clouds, Systems, and Services,

pages 48–61. Springer Berlin Heidelberg, 2010.

[100] Silaghi, G. C., ¸Serban, L. D., and Litan, C. M.

A time-constrained SLA negotiation strategy in

competitive computational grids. Future Gener-

ation Computer Systems, 28(8):1303–1315, oct

2012.

[101] Simao, J. and Veiga, L. Flexible SLAs in the

cloud with a partial utility-driven scheduling ar-

chitecture. In 2013 IEEE 5th International Con-

ference on Cloud Computing Technology and

Science. IEEE, dec 2013.

[102] Simao, J. and Veiga, L. Partial utility-driven

scheduling for ﬂexible SLA and pricing arbi-

tration in clouds. IEEE Transactions on Cloud

Computing, 4(4):467–480, oct 2016.

[103] Singh, A. and Viniotis, Y. An SLA-based re-

source allocation for IoT applications in cloud

environments. In 2016 Cloudiﬁcation of the In-

ternet of Things (CIoT). IEEE, nov 2016.

[104] SLA, C. Cloud service level agreement standard-

isation guidelines. European Commission, Brus-

sels, page 141, 2014.

[105] Snow, A. P. and Weckman, G. R. What are

the chances an availability SLA will be violated?

In Sixth International Conference on Networking

(ICN'07). IEEE, apr 2007.

[106] Snow, A. P., Weckman, G. R., and Gupta, V.

Meeting SLA availability guarantees through en-

gineering margin. In 2010 Ninth International

Conference on Networks. IEEE, 2010.

[107] Sommers, J., Barford, P., Dufﬁeld, N., and Ron,

A. Accurate and efﬁcient SLA compliance mon-

itoring. ACM SIGCOMM Computer Communi-

cation Review, 37(4):109, oct 2007.

[108] Sommers, J., Barford, P., Dufﬁeld, N., and Ron,

A. Multiobjective monitoring for SLA compli-

ance. IEEE/ACM Transactions on Networking,

18(2):652–665, apr 2010.

[109] Son, J., Dastjerdi, A. V., Calheiros, R. N., and

Buyya, R. SLA-aware and energy-efﬁcient dy-

namic overbooking in SDN-based cloud data

centers. IEEE Transactions on Sustainable Com-

puting, 2(2):76–89, apr 2017.

[110] Son, S., Jung, G., and Jun, S. C. An SLA-

based cloud computing that facilitates resource

allocation in the distributed data centers of a

cloud provider. The Journal of Supercomputing,

64(2):606–637, jan 2013.

[111] Stanik, A., Koerner, M., and Lymberopoulos, L.

SLA-driven federated cloud networking: Qual-

ity of service for cloud-based software deﬁned

networks. Procedia Computer Science, 34:655–

660, 2014.

[112] Starmer, C. Developments in non-expected util-

ity theory: The hunt for a descriptive theory of

choice under risk. Journal of Economic Litera-

ture, 38(2):332–382, jun 2000.

[113] Study Group 2, I.-T. Framework of a ser-

vice level agreement for ITU-t recommendation

e.860. Technical report, june 2002.

[114] Study Group 2, I.-T. Deﬁnitions of terms related

to quality of service for ITU-t recommendation

e.800. Technical report, Sep 2008.

[115] Takahashi, T., Kannisto, J., Harju, J., Heikkinen,

S., Silverajan, B., Helenius, M., and Matsuo, S.

Tailored security: Building nonrepudiable secu-

rity service-level agreements. IEEE Vehicular

Technology Magazine, 8(3):54–62, sep 2013.

[116] Tang, B. and Tang, M. Bayesian model-based

prediction of service level agreement violations

for cloud services. In 2014 Theoretical Aspects

of Software Engineering Conference. IEEE, sep

2014.

[117] Thangaraj, J., Mankar, P. D., and Datta, R. Im-

proved shared resource allocation strategy with

SLA for survivability in WDM optical networks.

Journal of Optics, 39(2):57–75, jun 2010.

[118] Trienekens, J. J., Bouman, J. J., and van der

Zwan, M. Speciﬁcation of service level agree-

ments: Problems, principles and practices. Soft-

ware Quality Journal, 12(1):43–57, mar 2004.

[119] Trygar, T. and Bain, G. A framework for service

level agreement management. In MILCOM 2005

- 2005 IEEE Military Communications Confer-

ence. IEEE.

[120] Van, H. N., Tran, F. D., and Menaud, J.-M. SLA-

aware virtual resource management for cloud in-

frastructures. In 2009 Ninth IEEE International

Conference on Computer and Information Tech-

nology. IEEE, 2009.

[121] Varela, M., Zwickl, P., Reichl, P., Xie, M., and

Schulzrinne, H. From service level agreements

(SLA) to experience level agreements (ELA):

The challenges of selling QoE to the user. In

2015 IEEE International Conference on Commu-

nication Workshop (ICCW). IEEE, jun 2015.

[122] Walter, J., Okanovi

c, D., and Kounev, S. Map-

ping of service level objectives to performance

queries. In Proceedings of the 8th ACM/SPEC

on International Conference on Performance En-

gineering Companion - ICPE '17 Companion.

ACM Press, 2017.

[123] Watson, B. J., Marwah, M., Gmach, D., Chen,

Y., Arlitt, M., and Wang, Z. Probabilistic per-

formance modeling of virtualized resource allo-

cation. In Proceeding of the 7th international

conference on Autonomic computing - ICAC '10.

ACM Press, 2010.

[124] Wu, L. and Buyya, R. Service level agreement

(SLA) in utility computing systems. In Grid and

Cloud Computing, pages 286–310. IGI Global.

[125] Wu, L. and Buyya, R. Service Level Agree-

ment (SLA) in utility computing systems. CoRR,

abs/1010.2881, 2010.

[126] Xia, M., Batayneh, M., Song, L., Martel, C. U.,

and Mukherjee, B. SLA-aware provisioning

for revenue maximization in telecom mesh net-

works. In IEEE GLOBECOM 2008 - 2008 IEEE

Global Telecommunications Conference. IEEE,

2008.

[127] Yan, J., Kowalczyk, R., Lin, J., Chhetri, M. B.,

Goh, S. K., and Zhang, J. An agent negotia-

tion approach for establishment of service level

agreement. In Computer Supported Coopera-

tive Work in Design III, pages 459–468. Springer

Berlin Heidelberg, 2007.

[128] Yaqub, E., Yahyapour, R., Wieder, P., Kotsokalis,

C., Lu, K., and Jehangiri, A. I. Optimal ne-

gotiation of service level agreements for cloud-

based services through autonomous agents. In

2014 IEEE International Conference on Services

Computing. IEEE, jun 2014.

[129] Zhao, J. and Subramaniam, S. QoT- and

SLA-aware survivable resource allocation in

translucent optical networks. In 2015 IEEE

Global Communications Conference (GLOBE-

COM). IEEE, dec 2015.