Combining SLA prediction and cross layer adaptation for preventing SLA violations

Abstract

Service-based Applications (SBA) are deployed in highly dynamic and distributed settings, where various parts of the constituent components -services and their infrastructure- are controlled by different third parties. In such a loosely coupled environment, adaptation capabilities are needed to manage deviations and unforeseen situations which might lead to negative consequences (e.g. contractual penalties). Current approaches either focus on cross-layer adaptation or the prevention of SLA violations. In contrast to this, the approach presented in this paper combines both. The paper presents an architecture as a generic framework for the management of arising problems during service execution. Multiple adaptation mechanisms are available to react on adaptation needs, acting on different layers of the SBA (including e.g. the composition layer and the infrastructure layer). The nal goal of the cross-layer adaptation capability is to avoid the violation of agreed Service Level (in SLAs) and thus ensure the bene ts of SBAs for both customers and providers.