AutoSLAM - Policy-based Management of Service Level Agreements in Dynamic Environments
This thesis investigates the problem of Service Level Agreement (SLA) management in diverse and dynamic service provisioning environments
such as the cloud and Everything as a Service (XaaS). It particularly focusses on SLA establishment
, which is a complex business process where consumers and providers, with varying and potentially conflicting preferences
, interact with one another in order to reach mutually acceptable agreements over the service usage terms and conditions. These interactions are governed by public interaction protocols
that define the participants’ communicative behaviour and are guided by private interaction strategies
that define their decision-making behaviour. However, the diverse, transient
nature of the multi-dimensional interaction context
can potentially impact the interactions during SLA formation and fulfilment. Therefore, we posit that it is not sufficient for service entities to use a static set of hard preferences and a single interaction protocol tightly coupled with one decision-making strategy for SLA establishment. Instead, they can benefit from flexible and expressive preference models, support for multiple interaction protocols and decision-making strategies, and the ability to adapt their interaction behaviour in response to changes in the interaction context. Having these capabilities, service entities can participate more efficiently and effectively in the quality-aware and quality-assured provision of services.
Given the unique characteristics of the SOA environment, the key requirements for automated SLA establishment are:
- Support for expressive preferences A service is normally characterised by multiple customizable attributes, with conflicting and dynamic preferences based on which trade-offs can be made. So one of the key requirements of the automated SLA establishment is the support for expressive and flexible preference statements.
- Support for multiple interaction protocols There is no standard protocol that can be used for SLA establishment in all scenarios and contexts and different parties can implement different protocols. Hence, an automated SLA management system should be able to support multiple interaction protocols.
- Support for multiple decision-making strategies An enterprise may have to use different decision-making strategies and algorithms depending upon the service provisioning context. Hence, it should support a wide range of decision-making strategies and algorithms, and be able to decide which strategy to use with which protocol when depending upon the contextual conditions.
- Support for simultaneous SLA interactions A service provider simultaneously provisions its service to a diverse set of consumers. It will have different high-level objectives it wants to achieve from each of these different negotiations. As a simple example, let us consider a startup company offering IaaS. If it is negotiating with a Fortune 500 company, its main business goal might be to procure a deal without worrying about profitability. On the other hand, if the consumer is a one-man company, the company may be interested in completing the negotiation as quickly as possible and getting the maximum profit from the deal.
- Context dependant decision making The negotiating parties are heterogeneous in open environments and the negotiation context can change constantly. The objectives of the negotiation for an organization will be different in the different contexts. Consequently, the specific strategy chosen is influenced by the contextual information available to the policy engine and the available policies so that the results of negotiation are consistent with the company's strategic goals and objectives.
refers to any situational information
that affects the negotiation process. It captures the states and conditions that are relevant to an enterprise during the process of negotiation. It can include information about the counterparts (size of the company, credit rating of the company, history of previous negotiations
etc), about itself (current load, current demand, availability of resources
etc) and the market in general. In open and dynamic environments like the cloud, the negotiation context is likely to change constantly. Depending upon the context, the goals of negotiations change, which means that different strategies will have to be used in different negotiation scenarios.
To address the above issues, we propose the AutoSLAM (Automated SLA Management) Framework
- an unified policy-based framework with corresponding models, mechanisms and tools for the automated establishment of SLAs in SOA environments. Its key features include:
- A formal preference model for specifying requirements and capabilities, and preferences over them in a flexible and expressive manner (in the form of preference policies)
- A formal choreography model for specifying the SLA choreography. SLA choreography enumerates the roles involved in the collaboration, the allowed interactions between them, the sequence and conditions under which the messages are exchanged, and the message contents i.e. it captures the Interaction Protocol, and
- A formal context-driven decision model for specifying the decision-making behaviour that guides the SLA interactions (in the form of decision rules)
- A physical model corresponding to the above formal models WS-SLAM is a generic policy specification language which extends the WS-Policy Framework and can be used for specifying the preferences, interaction protocols and decision-strategies in a declarative manner
- A policy-based orchestration model of SLA establishment, which enables flexible and adaptive management of SLA establishment through a loose coupling between strategic and communicative behaviour
- A reference architecture for the AutoSLAM runtime environment which can be easily implemented and extended
- A proof-of-concept prototype of the AutoSLAM middleware and a real-world application that uses it for the adaptive provisioning of computing resources on Amazon EC2
The key benefits of using the AutoSLAM Framework include:
- The AutoSLAM middleware gives service consumers and providers the flexibility to choose the interaction model that is most appropriate in a given context, while simultaneously participating in multiple concurrent SLA interactions using different interaction models.
- Policy-based orchestration of SLA choreography enables flexible and adaptive SLA establishment which can effectively deal with the diverse, transient, volatile and multi-dimensional nature of the interaction context. It does this by providing a loose coupling between the communicative and decision-making behaviour which allows entities to adapt their SLA interactions in response to changes in the interaction context. The adaptive action during an ongoing SLA interaction might be to:
- change some parameters of the current decision-making strategy
- change the current decision-making strategy, or
- change both the interaction-protocol and decision-making strategy
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: Agent-enabled Adaptive Management of Cloud Service Provisioning, Accepted for publication In 2015 IEEE/WIC/ACM International Conference on Intelligent Agent Technology (Accepted on 21/07/2015) (Acceptance Rate - 26%)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: Supporting Temporal Aspects of SLA Establishment in AutoSLAM Framework, 12th International Conference on Service Computing (SCC 2015), June 27 - July 2, 2015, New York, USA (CORE A -Accepted on 13/04/2015)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: Adaptive AutoSLAM - Policy-based Orchestration of SLA Establishment, In the Proceedings of the 11th International Conference on Service Computing (SCC 2014), June 27 - July 2, 2014, Anchorage, Alaska, USA (CORE A -Accepted on 26/03/2014)
- Mohan Baruwal Chhetri, Bao Vo and Ryszard Kowalczyk (2013). AutoSLAM - A Policy-based Framework for Automated SLA Establishment in Cloud Environments Accepted for publication in Concurrency and Computation: Practice and Experience, John Wiley & Sons (accepted on 18/09/2013) (CORE A)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: AutoSLAM - A Policy-driven Middleware for Automated SLA Establishment in SOA Environments, Accepted for publication in the Proceedings of the 9th International Conference on Service Computing (SCC 2012), 24-29 June 2012, Honolulu, Hawaii, USA (CORE A -Acceptance-Rate - 18%)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: Policy-Based Automation of SLA Establishment for Cloud Computing Services, In Proceedings of the 11th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID-12), Ottawa (Canada), 13-16 May 2012 (CORE A) (Accepted for publication on 30th January 2012 - Acceptance Rate - 27.5%)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: A Flexible Policy Framework for the QoS Differentiated Provisioning of Services, The 11th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID-11), Newport, California, USA, May 23-26, 2011 (CORE A) (Accepted for publication on 29th January 2011)
- Mohan Baruwal Chhetri, Bao Quoc Vo, Ryszard Kowalczyk: Policy-based Management of QoS in Service Aggregations, In Proceedings of the 10th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID-10), Melbourne (Australia), 17-20 May 2010 (CORE A)