Adaptive trade-off strategy for bargaining-based multi-objective SLA establishment under varying cloud workload

摘要

In cloud computing, a cloud management platform need to deal with three main aspects of the system: price, service performance, and resource utilization. They want to maximize profit while guaranteeing service performance agreed upon service level agreement (SLA), and increasing resource utilization. These multiple objectives of a cloud provider are difficult to be achieved individually since they are in a trade-off relationship (e.g., high resource utilization may lead to deterioration in service quality; expensive service price may lower resource utilization). This research focuses on the SLA establishment, which collectively considers business, service, and resource utilization aspects to achieve high profitability, guaranteed SLA, and efficient resource management. Thus, this paper proposes an adaptive negotiation mechanism for multi-objective SLA establishment under varying cloud workload. The proposed mechanism adaptively controls negotiation parameters, which represents preferences among multiple SLA issues under a trade-off, by analyzing workload trends. Using the proposed mechanism, a cloud system can shift on-peak load and alleviates SLA violations with flexible pricing. Consequently, the contributions of this paper include (i) the design of an adaptive negotiation mechanism for multi-objective SLA establishment, (ii) a guidance for determining SLA negotiation parameters for cloud pricing and resource management, and (iii) a demonstration that shows bargaining-based SLA establishment facilitates cloud resource management, and increases profit of cloud computing system. Empirical results conducted in a cloud testbed show that the proposed mechanism achieves higher performances than related approaches in terms of SLA violations and provider's profits.