Client Classification Policies for SLA Enforcement in Shared Cloud Datacenters

In Utility Computing business model, the owners of the computing resources negotiate with their potential clients to sell computing power. The terms of the Quality of Service (QoS) and the economic conditions are established in a Service-Level Agreement (SLA). There are many scenarios in which the agreed QoS cannot be provided because of errors in the service provisioning or failures in the system. Since providers have usually different types of clients, according to their relationship with the provider or by the fee that they pay, it is important to minimize the impact of the SLA violations in preferential clients. This paper proposes a set of policies to provide better QoS to preferential clients in such situations. The criterion to classify clients is established according to the relationship between client and provider (external user, internal or another privileged relationship) and the QoS that the client purchases (cheap contracts or extra QoS by paying an extra fee). Most of the policies use key features of virtualization: Selective Violation of the SLAs, Dynamic Scaling of the Allocated Resources, and Runtime Migration of Tasks. The validity of the policies is demonstrated through exhaustive experiments.