Conference PaperPDF Available

Service Level Agreement Management with Adaptive Coordination

Authors:
  • Whitestein Technologies AG, Switzerland

Abstract and Figures

Service Level Agreement Management in the telecom- munications domain consists of a set of mechanisms for provisioning and monitoring services according to require- ments given by either a customer or provider. This pa- per presents an approach to adaptive coordination between customers requesting service via an SLA, and the provider who owns and provisions network resources. We deploy a set of software agents capable of automating negotiation between these parties, while applying provider policy and controlling admission of service requests onto the underly- ing network infrastructure. An architecture supporting this approach is described, as is a prototype of the full system deployed on a network simulator.
Content may be subject to copyright.
Service Level Agreement Management with Adaptive Coordination
Dominic Greenwood and Giosu`
e Vitaglione and Lukas Keller and Monique Calisti
Whitestein Technologies AG
Pestalozzistrasse, 24
CH-8032 Z¨
urich, Switzerland
{dgr, gvi, lke, mca}@whitestein.com
Abstract
Service Level Agreement Management in the telecom-
munications domain consists of a set of mechanisms for
provisioning and monitoring services according to require-
ments given by either a customer or provider. This pa-
per presents an approach to adaptive coordination between
customers requesting service via an SLA, and the provider
who owns and provisions network resources. We deploy a
set of software agents capable of automating negotiation
between these parties, while applying provider policy and
controlling admission of service requests onto the underly-
ing network infrastructure. An architecture supporting this
approach is described, as is a prototype of the full system
deployed on a network simulator.
1. Introduction
In contemporary communication networks, many as-
pects of the interactions between service providers and their
customers are regulated by static and relatively inflexible
Service Level Agreements (SLA). Bound into the parame-
ters of these contracts are details such as the number of
links, network nodes (or access points) and the available ca-
pacity across end-to-end paths used to deliver content, per-
haps at using a certain pricing model. What is all too often
neglected is an account of the current network state and re-
source deployment balanced against stringent realtime con-
straints (resources being anything from physical network
equipment to available capacity across a link). As a conse-
quence networking performance and management tasks are
strongly constrained, since the options available to dynam-
ically accommodate offerings and balance the load across
the network resources is very limited or even impossible.
Accepted for the International Conference on Network-
ing and Services ( ICNS’06), July 19-21, 2006, Silicon
Valley, USA.
This naturally results in inefficient utilization of network
resources with suitably consequential losses in revenue.
The strong limitations of traditional approaches for ser-
vice provisioning and SLA Management 1(SLAM) are di-
rectly related to the lack of flexible and efficient meth-
ods/tools capable of proactively and dynamically support-
ing providers decisions. This is compounded by the diffi-
culties encountered by human operators when attempting to
consider all factors that influence the SLAM process and
make decisions in real-time. This complexity includes the
increasing number of actors in deregulated Telecom mar-
kets and the use of often heterogeneous technologies.
Considering these aspects, todays and tomorrows evolv-
ing network scenarios require a management solution that
uses both static and/or mobile software entities to collect
and analytically process network state information to sup-
port the decisions taken by human operators and, when suit-
able or necessary, directly effect changes in network com-
ponents.
We report on our solution to this requirement; a SLAM
System that directly employs software agent technology to
effect a proactive and flexible means of managing resources
and SLAs in IP QoS based networks. We describe how
agents are used to coordinate the decision process of de-
termining the value of SLA parameters between customers
and operators. We also show the use of a dedicated policy
server agent to control this decision process.
This SLAM System has been developed to software pro-
totype stage connected to a network simulator to evalu-
ate behavioral response under varying network conditions,
policies and customer requirements. This prototype, devel-
oped using the Living Systems R
Technology Suite2[3], is
briefly described.
The following section of the paper gives an overview of
our position on Service Level Agreement Management and
1See the TeleManagement Forum technical program on SLAM at
http://www.nmf.org/browse.asp?catID=1690
2For further information on LS/TS see: http://www.whitestein.com
/pages/solutions/ls ts.html.
the specific benefits brought about through the use of soft-
ware agent technology. In this section we also define our
notion of Service Access Control which governs the accep-
tance of a service provisioning request on the network in-
frastructure. Section 3 then describes our SLAM architec-
ture, including the agent system, coordination model and
policy model. Finally Section 4 provides a brief overview
of our prototype implementation before closing the paper
with conclusions.
2. Service Level Agreement Management
The basis for SLAM is the Service Level Agreement
(SLA) [2] which is a documented agreement between a ser-
vice provider and a service recipient and defines the basis of
understanding between the two parties for delivery of a ser-
vice. We define an SLA as containing clauses defining ser-
vice type, service level, QoS parameters such as delay, loss
and jitter, reservation information, pricing, etc. This SLA
is reified into an SLS [1], which is technical mapping of
the SLA that can be used by the Service Admission mecha-
nisms of network infrastructure.
The objectives of our work, are to provide an agent-based
framework for efficient and flexible Service Level Agree-
ment Management in telecommunication networks. This
is contingent on the capacity to negotiate and re-negotiate
SLAs, to warn the customer of possible violations of the
service guarantees and also to be able to reliably estimate
the effect of SLA admissions or tear downs on the network
properties in a short and midterm perspective in order to en-
sure a sound basis for negotiation. The task of estimating in
advance whether the QoS guarantees of a service instance
can be kept during the entire lifetime of that service is man-
aged by Service Access Control (SAC) to clearly separate it
from Call Admission Control (CAC) which denotes the be-
haviour of routers to reject certain calls in case where they
are overloaded.
Software agent technology [6] [3] [4] is a means of
creating autonomous, intelligent and social software as-
sistants capable of supporting human decision-making. An
agent comprehends and interacts with its environment and
components present within that environment, e.g., humans,
agents, services, etc. There is some evidence in the litera-
ture of software agents being applied to the SLAM domain,
including [5] which deals with management and monitor-
ing of SLAs.
Software agents in SLAM provide the means to support
and automate process and decisional aspects of end users
and/or service providers behaviour. In this respect they help
to:
Reduce the complexity of tasks including time con-
strained negotiation of SLA parameters, pricing sched-
ules and even dynamic network re-configuration.
Achieved by implementing knowledge processing and
analytical techniques that support and optimize the hu-
man decision making process. Complexity of control
is pushed into software agents: controlling or modi-
fying agents is easier and cheaper than changing the
network and management stack.
Provide a scalable network management approach
with flexible task assignment across a population of
agents
Offer value added services, such as the integration of
activities performed during SLA set up with other net-
work management services, or the personalization and
custom-tailoring of some tasks for which there is now
no means of differentiating the way end users and ser-
vices are monitored and maintained.
3. SLAM Architecture
The general architecture of the SLAM system is shown
in Fig. 1. As mentioned, the key features of our approach to
SLAM is the use software agents to efficiently and flexibly
negotiate the specific details of SLAs, apply operator policy
overlays to SLA decisioning, provision and monitor desired
and specified Quality of Service (QoS) levels, automatic re-
porting and optimal deployment of network resources ac-
cording to customer history models.
Figure 1. SLAM Architecture
From Fig. 1 we can see that SLAM consists of two per-
spectives. That of the customer and that of the provider.
Each customer is represented by a customer agent, often lo-
cated on a user device or item of network terminating equip-
ment. These agents coordinate with the network provider’s
service agents to agree on the parameters of SLAs defining
some service provisioning. The provider uses policy agents
to govern the decisioning aspect of the customer coordina-
tion and edge router agents to control Service Level Ac-
cess (SAC) onto the core network. SAC is the mechanism
by which a service is accepted or not for provisioning on a
network infrastructure and is the key mechanism for main-
taining preferred levels of QoS. Whenever a new request
for service made (via a Service Level Specification), SAC
is reponsible for verifying whether the conditions of the re-
quest set can be satisfied with currently available network
resources.
3.1. Agent System Model
The SLAM System is composed of several interacting
software agents, as indicated in Fig. 2. Customer agents
are unique in that they belong to the customer perspective,
i.e., are essentially operated by customers from their local
devices. The remainder of the agent classes are owned and
operated by the service provider and reside within the enter-
prise boundary. Each class of agent can exist in multiplicity,
as required according to topological, load or other distribu-
tion constraints.
Figure 2. SLAM Agent System Model
The agent classes are:
Customer Agents. These agents reside on customer de-
vices and act to support their human owner in negotiat-
ing SLA setup and dynamic change according to speci-
fied requirements and constraints. In simulation mode,
the customer agents can also mimic the expected be-
haviour of real customers by generating SLA requests
according to a variety of selected algorithms. Thus
simulated contracts can be established with static or
dynamic transient characteristics that describe service
requirements and financials for individual or clusters
of customers. In this way, the effects of different load
conditions of groups of users with different user pro-
files can be simulated and evaluated.
Service Agents. These agents are the operational heart
of the SLAM System and logically reside between
the service consumers (customers) and the service
provider. Their primary responsibility is to negoti-
ate the specific parameters contained within an SLA.
Typical parameters include service type, service level
(equates to quality class), pricing and if necessary fine
tuned variables for bandwidth, delay, loss, jitter, etc.
Service Agents must also verify all negotiated SLAs
with the policy agents to ensure that no contracts or
operational policies are violated.
Policy Agents. The Policy Agents are responsible for
administering provider policy. Each policy agent con-
tains a filter chain which contains one or more sequen-
tially connected filters. An incoming proposal (i.e., an
agreement formed between customer and provider for
the delivery of a service), is passed through each filter
in turn, each of which has veto power to reject the pro-
posal if certain conditions are not met. Filters can be
added or removed at runtime and loopbacks triggered
if necessary. Some examples of filters include verifi-
cation that established customer contracts are not vio-
lated by a negotiated SLA, and filters that ensure that
business rules defined by the provider are observed and
applied appropriately to service provisioning.
Edge Router Agents. The Edge Router Agents reside
at the network edge and are responsible for interfacing
between the SLAM service agents and the underlying
network infrastructure. Their primary task is to man-
age Service Admission Control. This is the process
by which a service provisioning specification (i.e., the
SLS, transcribed from the original SLA), is checked
against available network resources and either admit-
ted or rejected accordingly. The edge router agents use
operational constraints that define the levels of accept-
able performance that must be maintained on the net-
work infrastructure, e.g., acceptance thresholds to en-
sure a desired level of QoS is maintained across the
network provisioning.
GUI Agent. This agent simply acts as an interface be-
tween the GUI front-end and the SLAM Core agents.
NetSim Agent. This agent is responsible for control-
ling the underlying network simulation over which
the SLAM System can be tested. The simulator is
a detailed, purpose built, system that uses a standard
M/M/1/k queuing model for representing both routers
and links. The simulator contains functional models
for Diffserv, RSVP and MPLS. We consider any fur-
ther description of this model to be out of scope for
this paper.
3.2. Primary SLAM Process
The high-level operational sequence of the SLAM Sys-
tem is as follows:
1. A customer agent formulates SLA detailing service
provisioning request and sends to a provider service
agent.
2. A service agent accepts the SLA request, converts into
an SLS (Service Level Specification) and forwards the
request to an edge router agent having applied policy
from the policy agent as appropriate.
3. The receiving edge router agent and service agent ne-
gotiate the specific terms of the service provision-
ing based on SLA requirements and network resource
availability.
4. Proposals are returned to the customer agent, who can
accept or reject offers. Rejections typically trigger re-
negotiation.
3.3. Adaptive Coordination Model
The SLAM System employs several mechanisms to in-
crease the efficiency of the service provisioning by fully ex-
ploiting the available capacities of the network. Most of the
mechanisms are based on a better coordination of the po-
tentially conflicting interests of customers and providers.
Although we can assume that in a longer term perspec-
tive, every provider essentially aims at increasing its rev-
enue, its immediate goals can be manifold: optimizing net-
work utilization, reducing operational costs, and maximiz-
ing customer satisfaction (minimizes the SLA preemption,
violation and rejection rates) in order to retain customers
and reduce churn rate.
Customers simply want the best possible service for the
cheapest possible price. However, there are many different
classes of customers, which are best satisfied with a differ-
ent balance between prices and quality of the service.
SLAM performs service access negotiation taking into
account the status of the network, and exploiting availability
of resources that would remain unused with a static network
behaviour.
A number of interaction types have been identified that
increase the efficiency and flexibility of the system:
Sequential Coordination. If the service properties can
not be guaranteed as requested by the customer, the
system (in particular the Service Agent) sends alter-
native proposal(s) that can be satisfied by the current
status of the network and whose content is as close to
the conditions of the original request as possible. If
such proposal is rejected by the customer agent, a lim-
ited number of additional proposals can be formulated
by the system and sent to the customer.
Concurrent Coordination. Service requests coming
from multiple users are no longer treated individually
but several of them arrived in the same time interval
are evaluated together. In case of a lack of resources,
those that seem to be the most beneficial are treated
preferentially. Concurrent coordination allows better
improved optimisation of network resources.
Deferred Service Setup. If the customer agent consid-
ers as not satisfying the received offers, he can decide
to register a standing request at the Service Agent to-
gether with an expiration date and an upper price limit.
If within its lifetime the system can provide the re-
quired quality to the specified costs, the user will be
notified. Another related mechanism is the advanced
reservation of resources, in the sense that services are
not set up immediately after the conclusion of the SLA
agreement, but at some selectable moment in the fu-
ture. This would require a strict scheme for resource
reservation or a predictive network state estimations
mechanism.
On-line Re-negotiation. In some cases, re-negotiation
of a service being provided might free resources to
satisfy multiple service requests, possibly improving
the overall efficiency. In this kind of interaction, the
Service Agent initiates a negotiation with a Customer
Agent who is being provisioned a service. This inter-
action can result into a new agreement with modified
service parameters or even service type.
Push Service Offering. The Service Agent sends Spe-
cial Offers to the Customer Agent according to the sta-
tus of the network. The system could automatically, or
assist to, estimate the impact on providers revenue and
network usage of special offers like temporary price
reductions or combination-offers.
To date we have concentrated our efforts on the imple-
mentation of a viable sequential coordination mechanism.
Some of the more advanced concepts that have not been
implemented will be discussed at the end of this chapter. A
schema describing the agent coordination interactions be-
fore the setup of a service is shown in Fig. 3:
The Service Agent maintains a map containing default
traffic parameters (DSCP, preemption priority, RSVP pool)
rg
:CustomerAgent sa :ServiceAgent era
:EdgeRouterAgent
alt
[modify]
[accept]
[reject]
loop
[!rejected && !offerable &&!setup]
loop
[!rejected && !accepted]
alt
[rejected]
[setup]
[offerable]
paa
:PolicyAdministrationAgent
request:= getNextRequest
initRequest(request)
ServiceReqFragment(tempy)
sls:= setDefaultTrafficParams(request)
CheckReqFragment(currentSLS)
CheckDecFragment(cacResult)
request:= createAlternative(request,cacResult)
CheckReqFragment(currentSLS)
CheckDecFragment(cacResult)
ValidateReqFragment(params,action)
PAAReplyFragment(params,action)
RejectedFragment
SetupConfFragment(setupServ)
SetupConfirmFragment()
SetupConfFragment()
SLAList:= createSLAProposal(request)
EvaluateProposalFragment(SLAList)
ServiceReqFragment(tempy)
CustomerDecFragment(customerDec)
SetupConfFragment(setupServ)
SetupConfirmFragment()
SetupConfFragment()
CustomerDecFragment(null)
RejectedFragment()
Figure 3. SLAM Coordination Interactions
for the different service types. Upon reception of a new re-
quest, the Service Agent assigns it the default parameters
and submits it to the Edge Router Agent for SAC. For all
those LSPs to which the request could be assigned, the SAC
procedure on the Edge Router Agent now returns precise in-
formation on the quality of the flows of the tested traffic pa-
rameter configuration (CACData cacResult) and for the oth-
ers it indicates the reason why the request can not be routed
over them. The Service Agent uses then this information to
decide on the next actions it will take. When receiving the
SAC results, the Service Agent can opt for different actions:
Setup. If, with the default parameter settings, there is a
path where the quality is good enough for the required
service, it can be setup directly and the customer is
notified of its availability.
Offer. If it is not sure whether the quality of the pro-
posals is good enough or if wished by the user, the
Service Agent selects a subset of the Edge Routers pro-
posals and offers them to the customer. Depending on
the providers policy, the offers may be editable (of-
ferMod), in which case the customer can modify the
service requirements of an SLA and resubmit it to the
SA, or non-modifiable. In the latter case (offer), the
customer can only either accept one of the proposals
or reject the whole set and stop the interaction.
Continue. Considering the results received for previ-
ously tested parameter combinations, it can test the
quality and resource availability for different than the
default traffic parameters until it has either found one
(or several) combination(s) that match(es) the services
requirements and can be proposed to the customer or
until all meaningful combinations have been tried.
Reject. If the Service Agent is convinced, that the re-
quest can currently not be served, it is rejected.
Which of these actions will be taken under which pre-
cise conditions depends essentially on the configuration of
the Service Agent. When proposals are sent back to the
customer, the customer can either (1) Select the best and
ask the Service Agent to set it up (accept). If there is no sat-
isfying offer, he can (2) Reject them and stop, such that the
coordination ends without service setup (stop) or, if allowed
by the provider, (3) Specify other quality requirements and
resubmit the request to the Service Agent to obtain more
proposals (continue).
3.4. Policy Enforcement
Whereas the conformance of the particular service pa-
rameters to the customers requirements is checked by the
service agent, it is the responsibility of the policy agent to
enforce the providers goals. Since the service agent double
checks all important decisions first with the policy agent,
the latter is able to control the behaviour of the entire SLAM
system.
The definition and enforcement of declarative generic
administration policies is still a research topic and out of
scope for this paper. Therefore we focus on the design
and implementation of a set of simple procedural pluggable
policies. Policy enforcement is performed through the in-
teraction between the service agent and the policy agent.
Following interactions with a customer agent and an edge-
router agent, a service agent evaluates the SAC results, and
proposes the next action to be taken, i.e. further proposals to
the customer agent. These are passed to the policy agent to
obtain approval to proceed with the negotiation. The policy
agent can send back a subset of the proposals to the ser-
vice agent, imposing constrains according to the goals of
the provider expressed in the policies.
3.4.1 Policy Filter Chain
The policy agent is composed by a chain of policy filters
which allow variable composition of policies.
Provider’s business goals are mapped to one or more pol-
icy filters, that take the proposal as input, check its confor-
mance to the filters rules and if necessary adapts the propos-
als content accordingly. Depending on the nature of the goal
in case of offer and setup, the filter can attribute a scoring
value to each proposal that indicates how well the configu-
ration respects the filters goal. Above a certain threshold it
can veto the proposal, which is discarded.
Filters can be chained, such that the output of a first fil-
ter serves as input of a second one, the final output will be
submitted to and executed by a service agent. If a proposal
is vetoed by a filter, it is not even passed for evaluation to
the following filters in the chain (sine qua non conditions).
The filters can also change the proposal sent by the SA.
If, for example, a service agent proposes to offer a cer-
tain set of service configurations to the customer, the filters
on the policy agent can disregard the advice and decide to
change the quality provided according to a certain policy.
Note that if some filter changes the action, the new action
proposal must re-traverse the filter chain from the beginning
to assure its conformance with all the providers goals.
4. Implementation of the SLAM System
The SLAM System is implemented using the Living
Systems R
Technology Suite (LS/TS). This is a commer-
cially available, industry-strength software agent platform,
compliant with both J2SE and J2EE and with extensive
Eclipse based development and deployment tool support.
All SLAM agents are implemented in Java as LS/TS
agents with Petri net based behavioral logic. Communi-
cation is directly supported by the agent platform and the
agents may be distributed across multiple network devices.
The LS/TS administration tool allows the entire agent popu-
lation, databases, interactions and events to be monitored in
real time with intervention policies deployable as required.
A snapshot from the GUI front-end of the prototype
SLAM System is shown in Fig. 4.
Figure 4. SLAM Implementation GUI
This snapshot shows a network topology that can be
modified in real-time, with data on the right concerning the
particular network in live simulation and selected LSP (La-
bel Switched Path). On the left are several charts reporting
on performance of the simulation in terms of selected para-
meters such as active SLAs, the jitter, loss and delay associ-
ated with any given LSP or Node in the network simulator,
and behavior of the individual agents during decision and
coordination interaction.
5. Conclusions
This paper has presented an approach to achieving adap-
tive coordination between customers of telecommunica-
tions service providers and their providers, expressed using
SLAs. The reported system consists of a set of interacting
software agents that automate negotiation between the par-
ties, while allowing the application of policy controls, to
control service admission onto an underlying network in-
frastructure. The current prototype implementation can be
deployed and integrated within network management plat-
forms. A future version of the system will consist of agents
that use of the new Goal Based Planning feature of Liv-
ing Systems R
Technology Suite to enhance their behavioral
characteristics.
References
[1] D. Goderis et al. Service level specification semantics,
parameters and negotiation requirements. IEEE Inter-
net Draft, 2001.
[2] John Evans and Clarence Filsfils. Deploying diffserv at
the network edge for tight slas, part 2. IEEE Internet
Computing, 8(2):61–69, 2004.
[3] Giovanni Rimassa, Monique Calisti, and Martin Kern-
land. Living systems R
technology suite. Whitestein Se-
ries in Software Agent Technologies Systems: Software
Agent-Based Applications, Platforms and Development
Kits, 2005.
[4] I. Trencansky and R. Cervenka. Agent Modelling Lan-
guage (AML): A comprehensive approach to modelling
mas. Informatica, 29(4):391–400, 2005.
[5] F. De Turck, S. Vanhastel, P. Backx, B. B. Duyburgh,
and P. Demeester. Design of a generic architecture for
service management and monitoring of service level
agreements through distributed intelligent agents. In
Proceedings of IEEE Intelligent Network Workshop,
pages 50–57, 2001.
[6] Michael J. Wooldridge. Introduction to Multiagent Sys-
tems. John Wiley & Sons, Inc., New York, NY, USA,
2001.
... Hence, it very significant to tradeoff, the performance, and energy; to meet the SLA, the energy consumption must be reduced [15]. In the service level system, the expectation among the customer and the cloud provider for the delivered services like, responsibilities and penalties of the violation arise in the cloud parties are specifically designed in a document termed as service level agreement (SLA) [16][17][18]. Certain service level objectives (SLO), namely data throughput, available percentage, and the response time of service are defined in SLA for measuring the condition provided by the service. The SLO are integrated together to meet the expectation among both cloud parties and it significantly varies based on the outsourced data or the applications [16]. ...
Article
Full-text available
In the cloud service system, the expectation between the customer and the cloud provider for the delivered services like, responsibilities and penalties of the violation arise in the cloud parties are specifically designed in a document named as Service Level Agreement (SLA). However, violation in the Service Level Agreement is a major challenge in the research community of SaaS cloud providers. Hence, an effective violation detection method named Chicken Spider Monkey Optimization-based Deep Belief Network (ChicSMO-based DBN) is proposed in this research to perform the violation detection mechanism in the cloud environment. The proposed Chicken Spider Monkey Optimization-based Deep Belief Network effectively computes the scores, like User Privacy Score (UPS), Service Level Agreement score (SLA), and Cloud Service Provider score (CSP) based on the fitness value. The spider monkeys forage the food by evaluating the distance from the food, and update their position towards the food source. The feature scores are computed by updating the position of local and global leader group members, and perform the violation detection process using the Deep Belief Network classifier. The performance enhancement attained by the proposed algorithm in terms of the evaluation metrics, like accuracy, coverage rate, and F-value which acquire with the values of 0.862, 0.944, and 0.889 with the transaction data of 1000, respectively. It increases the reliability to address the QoS guarantee issues and increase the confidence level of the prediction model.
... In general, management of SLAs has been researched out in the past couple of years resulting mainly in the definition of languages and standards for the specification of SLAs [9,24]. Other work in this line of research particularizes the management of SLAs to specific domains [11,20,63]. A great deal of effort has been put in devising techniques and machinery for monitoring service level agreements [2,12,14,21]. ...
Article
Full-text available
As service-orientation gathers traction in enterprise engineering, more and more business processes and capabilities alike are wrapped up and viewed as services in today’s digital enterprises. Digital servitization is neither a rebranding nor a syntax sugar in enterprise engineering. It opens the door to innovative business models, novel system architecture, to name just a few. Central to service provisioning is the notion of commitment that captures contractual agreements between the provider and the consumer of the service in question, describing not only computer-interpretable factors (e.g., reliability, payment), but also business rules (e.g., service penalty) of the commitment. Widely known as the service level agreement, this building block of service-oriented enterprise engineering is pushed towards virtual when it comes to digital servitization. In this article, we propose a descriptive framework for digital enterprise services together with devised techniques for the alignment and reinforcement of the service level agreement. The relevance and applicability of the multi-perspective approach to service engineering are among our findings, which are illustrated using a few examples and a real-life case-study.
... An SLA is a legal contract which includes service obligations, deliverability, service objectives and service violation penalties [4][5][6]. An SLA is not only used to measure the performance of the provider, it also helps to resolve disputes regarding consumer duties [7]. ...
Article
Full-text available
In cloud computing, service level agreements (SLAs) are legal agreements between a service provider and consumer that contain a list of obligations and commitments which need to be satisfied by both parties during the transaction. From a service provider's perspective, a violation of such a commitment leads to penalties in terms of money and reputation and thus has to be effectively managed. In the literature, this problem has been studied under the domain of cloud service management. One aspect required to manage cloud services after the formation of SLAs is to predict the future Quality of Service (QoS) of cloud parameters to ascertain if they lead to violations. Various approaches in the literature perform this task using different prediction approaches however none of them study the accuracy of each. However, it is important to do this as the results of each prediction approach vary according to the pattern of the input data and selecting an incorrect choice of a prediction algorithm could lead to service violation and penalties. In this paper, we test and report the accuracy of time series and machine learning-based prediction approaches. In each category, we test many different techniques and rank them according to their order of accuracy in predicting future QoS. Our analysis helps the cloud service provider to choose an appropriate prediction approach (whether time series or machine learning based) and further to utilize the best method depending on input data patterns to obtain an accurate prediction result and better manage their SLAs to avoid violation penalties.
... The existing frameworks for SLA management only define the format and types of messages that can be exchanged during the negotiation between service providers and service consumers. Some projects particularize the management of SLAs to specific domains, such as military, database management, or information systems [8,13,25]. Garvin proposed a quality management grid featuring a total of eight dimensions including performance and reliability [11]. ...
... The existing frameworks for SLA management only define the format and types of messages that can be exchanged during the negotiation between service providers and service consumers. Some projects particularize the management of SLAs to specific domains, such as military, database management, or information systems [8,13,25]. Garvin proposed a quality management grid featuring a total of eight dimensions including performance and reliability [11]. ...
Conference Paper
Business services arguably play a central role in service-based information systems as they fill in the gap between the technicality of Service-Oriented Architecture and the business aspects captured in Enterprise Architecture. Business services have distinctive features that are not typically observed in Web services, e.g. significant portions of the functionality of business services might be executed in a human-mediated fashion. As such, service level agreement (SLA) should be described as a mixture of human-mediated functionality (e.g., service penalty) and computer-interpretable measurement (e.g., reliability, payment). In this paper, we propose a formal framework for reasoning about the SLAs from the perspective of services bundling – the practice of innovatively organizing business services into a bulkier service offering that creates new values. Specifically, we (a) represent multi-level SLA of a business service in terms of service reliability, payment and penalty using the mathematical structure of semiring; (b) provide formality for aggregating SLAs of the constituent services that make up the service bundling; (c) make multi-level SLAs of a bundled service technically comparable. The main contribution of this work is a machinery for handling a large number of SLAs generated through services bundling, allowing to the service consumers to pick up the right service offering according to their preference.
... Multi-agent based systems are currently being used in wide variety of applications, ranging from comparatively small systems for personal assistance to open, complex, mission critical systems for industrial applications [8] . An example of the application of multi-agent systems includes system diagnostics [9] and network management [10] which needs to be monitored and managed in real time. Multi-agent systems are also used both for the management of distributed networks and for the realization of advanced telecommunication services [11,1] . ...
Article
Since it is not easy for network administrators to monitor nodes in a network environment manually due to the physical movement from one node to another; and it is even worst with the trend of the increase in nodes in a network environment due to Web based Service Oriented Applications (SOA) that runs on the networks. Monitoring of nodes in terms of detecting faulty network cables, detecting of nodes that is not supposed to be part of a particular network, and even the shutting down of hundreds of nodes has become a nightmare for most network administrators. The static software solutions and dynamic single task software for network management developed has not really helped matters on the network administrators end. These worries gave birth to this research work, of designing a multi-agent based system to handle different task for the sole benefit of the network administrator and not the network environment. The system is effective and efficient, and it is recommended for practical usage.
... The expectations of providers and costumers of a cloud service including the penalties considered for violations are all documented in the Service Level Agreement (SLA) [68,71,162]. Considering SLA, energy management techniques are categorized into two groups, namely SLA-Aware and SLA-Agnostic approaches (as shown in Figure 2.14). ...
Thesis
Full-text available
C LOUD enables access to a shared pool of virtual resources through Internet and its adoption rate is increasing because of its high availability, scalability and cost effectiveness. However, cloud data centers are one of the fastest-growing energy consumers and half of their energy consumption is wasted mostly because of inefficient allocation of the servers resources. Therefore, this thesis focuses on software level energy management techniques that are applicable to containerized cloud environments. Containerized clouds are studied as containers are increasingly gaining popularity. And containers are going to be major deployment model in cloud environments. The main objective of this thesis is to propose an architecture and algorithms to minimize the data center energy consumption while maintaining the required Quality of Service (QoS). The objective is addressed through improvements in the resource utilization both on server and virtual machine level. We investigated the two possibilities of minimizing energy consumption in a containerized cloud environment, namely the VM sizing and container consolidation. The key contributions of this thesis are as follows: 1. A taxonomy and survey of energy-efficient resource management techniques in PaaS and CaaS environments. 2. A novel architecture for virtual machine customization and task mapping in a containerized enterprise cloud environment. 3. An efficient VM sizing technique for hosting containers and investigation of the impact of workload characterization on the efficiency of the determined VM sizes. 4. A design and implementation a simulation toolkit that enables modeling for containerized cloud environments. 5. A framework for dynamic consolidation of containers and a novel correlation-aware container consolidation algorithm. 6. A detailed comparison of energy efficiency of container consolidation algorithms with traditional virtual machine consolidation for containerized cloud environments.
Article
Cloud-based systems involve the management and control on the level of delivered service, which is crucial in order to maintain service continuity and customers’ trust for the service provided. Predefined service level agreement (SLA) is normally used as the main element in managing the cloud-based system. However, such SLA basically cannot be adjusted during operations. Current studies on SLA adjustment are mainly focused on modifying the agreement before commencing operations. The cloud environment is, however, dynamic and requires occasional review of the SLA parameters to manage expectations. Thus, to maintain service continuity that is acceptable to all parties, SLA adjustment is necessary during service operations. This paper presents a real-time and proactive SLA renegotiation model to support the dynamic nature of cloud-based systems. Renegotiation is shown for four different scenarios based on parameter weightage. In order to achieve real-time decision, a multi-offer generation approach is used. A novel mechanism to detect and predict service violation is proposed to ensure proactive renegotiation. Simulation is performed to verify the model. In the simulation, the model is found to be effective as it is able to generate multi-offers in a single renegotiation round and reduce the impact of service violations during service operations.
Conference Paper
Full-text available
With the rise of Internet of Things, end-users expect to obtain data from well-connected smart devices and stations through data services being provisioned in distributed architectures. Such services could be aggregated in a number of smart ways to provide the end-users and third-party applications with sophisticated data (e.g., weather data coupled with soil pollution), resulting in a growing number of service offerings to be requested. Service offerings that have been shortlisted for a certain data request (e.g., rainfall in a particular farming site) need to be ranked according to the end-users’ preference. Service level agreements, i.e., the mutual responsibilities between the service provider and its consumers, address this sort of preference. Unfortunately, provisioning quality-aware services under this term still stays on the sidelines. In this paper, we propose a service architecture where the service level agreements shall be: (i) accumulated overtime on IoT service transactions; (ii) compiled when aggregating IoT services; (iii) used as a ranking criterion for suggesting IoT service offerings.We demonstrate our machinery in the services provisioning of agricultural datasets taken from a farming site of the Mekong Delta in Vietnam.
Chapter
Agents and agent-based approaches are an active research topis in artificial intelligence and expert systems.
Article
Full-text available
The Agent Modeling Language (AML) is a semi-formal visual modeling language for specifying, modeling and documenting systems that incorporate features drawn from multi-agent systems theory. It is specified as an extension to UML 2.0 in accordance with major OMG modeling frameworks (MDA, MOF, UML, and OCL). The ultimate objective of AML is to provide software engineers with a ready-to-use, complete and highly expressive modeling language suitable for the development of commercial software solutions based on multi-agent technologies. This paper presents an overview of AML. The scope of the language, its structure and extensibility mechanisms are discussed, and the core AML modeling constructs and mechanisms are introduced and demonstrated by examples.
Conference Paper
Full-text available
We present a generic architecture for service management and monitoring of service level agreements (SLA). Special attention has been devoted to the efficient management of computational resources. The usage of an intelligent agent platform (IAP) with this respect is motivated. The case study of an adaptive distributed caching architecture is presented and the agent for adaptive distributed data caching is detailed. Furthermore, a sample scenario is provided to illustrate the different stages in the data caching process and how distributed intelligent agents are used in favor
Chapter
This chapter presents and discusses the Living Systems Technology Suite, LS/TS, a solution for the development and deployment of products and systems based on software agent technology and autonomic computing. LS/TS comprises a software development methodology and a Java-based agent platform with development tools. The focus of this paper is on the LS/TS agent platform: the concepts, API and development tools that support the design and implementation of multi-agent systems are described and discussed. This chapter also lists a few significant challenges that a middleware for multi-agent systems has to face, and also shows how each one of them is addressed by the LS/TS agent platform.
Article
In the second of a two-part series, the authors review industry best practices for designing, validating, deploying, and operating IP-based services at the network edge with tight service-level agreements (SLAs). Specifically, they present a case study that shows how Diffserv can be deployed to achieve these SLAs.
Article
In the first of a two-part series, we review industry best practices for designing, validating, deploying, and operating IP-based services at the network edge with tight service-level agreements (SLAs). We describe the important SLA metrics for IP service performance and discuss why Diffserv is the preferred technology to achieve these SLAs.