Ouantitative Management Of Database System Workload

The workload of contemporary database management systems (DBMSs) primarily stems from the network applications. With the rapid development of network applications, DBMSs' workloads become more and more diverse and complex. Different workloads (or network applications) have different Service Level Objectives (SLO). Meeting the workloads'SLO has become one of the important tasks of DBMSs. SLOs generally include two requirements, namely the workload importance and the performance objective. Quantifying workload importance means managing workloads towards their performance objectives according to their workload importance. It is an important function of autonomic DBMSs. As a high-level concept, workload importance cannot be precisely translated into precedence, a low-level construct. It is important and challenging to find a technique to precisely translate workload importance into workload management.This paper studies the approach by which workload importance can be managed quantitatively under the autonomic workload management framework (AWMF), in other words, translating workload importance into a real number, which can be used as a parameter in allocating resources to meet their performance objectives. The approach assigns an importance level to each workload based on their importance to business, and allocates resources accordingly. The paper introduces the autonomic workload management framework, AWMF, and proposes the techniques for quantifying workload importance and adjusting the workload importance level. The technique for quantifying workload importance map workload importance into a real number called workload importance level, which can be used as a parameter in allocation resources. The technique for adjusting workload importance level allows AWMF adjusting workload importance level without violating the sequence of workload importance, which gives AWMF a large space for allocating resources.