Goal-Driven Autonomous Database Tuning Supported by a System Model

Fast growing volumes of data, typically for today’s database applications, require efficient and low-cost management to guarantee reliable processing and high throughputs. Although administrative costs for skilled DBAs are noticeable, it is often done manually due to the lack of available online tuning utilities. Typical operating goals for a database system comprise throughput, resource utilization, memory footprint, and due to recent efforts in saving energy also the green factor of a system. However, such high-level goals conceal the entire system complexity but help to address the essential cost factors. With the help of an extendable system model, describing fix and variable input parameters, state, runtime behavior, and workload correlations of parameters and system components are disclosed. Based on that system model, goalspecific analyses should be possible to tune the system online or at least to recommend actions needed to fulfill the goals.