5 CONCLUSIONS AND FUTURE RESEARCH
5.1 Summary
In this paper, we present a systemic study of workload management in DBMSs. We surveyed workload management systems implemented in today’s commercial DBMSs and techniques proposed in the recent research literature. We propose a taxonomy of workload management techniques to classify workload management techniques and identity the techniques employed in a workload management system. The taxonomy categorizes workload management techniques into four major technique classes, namely workload characterization, query admission control, query scheduling and query execution control technique classes. In a workload management technique class, the techniques may be further divided into subclasses based on their distinct technique mechanisms. We also introduce the underlying principles of workload management technology used by today’s commercial DBMSs, which are outlined as defining performance objectives for arriving queries based on a given SLA, identifying the arriving queries present on a data server, and imposing controls on the queries to manage their behaviors in order to achieve the performance goals.
In the taxonomy of workload management techniques, we show that the typical technique used in the workload characterization technique class is workload definition associated with resource allocation. The typical technique used in the query admission control techniques class is setting thresholds for queries. The typical techniques used in the query scheduling technique class are managing query waiting queues and query restructure, and in the query execution control technique class, the typical techniques are query suspension, resource reallocation and query kill, in which the query suspension type techniques can be further divided into query throttling and query suspend-and-resume subtypes.
