Poster Session: Improved Buffer Size Adaptation through Cache/Controller Coupling

Database workloads seldom remain static. A system tuned by an expert for the current environment, might not always remain optimal. To deal with this situation, database systems have been incorporating self tuning features. An important component is self tuning memory or bufferpools. These often work on a feedback loop and take some time to converge to an optimal state. T)\c transition to the optimal state could be accelerated if future access patterns could be taken into account when making decisions on the bufferpool size. In this paper, we describe a caching algorithm for scans on bufferpools. which keeps track of ongoing scans and the state of each scan. T)\e proposed algorithm results in a better hit ratio and. more importantly, provides a way to predict future access patterns. This property is beneficial for providing feedback to a self-tuning memory controller to make better allocation decisions.