Batched disk scheduling with delays

Abstract

One of the important performance enhancing capabilities of modern disk drives, is the ability to permute the order of service of incoming I/O requests in order to minimize total access time. Given a batch (set) of I/O requests, the problem of finding the optimal order of service is known as the Batched Disk Scheduling Problem (BDSP). BDSP is a well known instance of the Asymmetric Traveling Salesman Problem (ATSP), in fact it has been used as one of a few principal test cases for the examination of heuristic algorithms for the ATSP, [4], [12]. To specify an instance of BDSP amounts to a choice of a model for the mechanical motion of the disk and a choice of locations and lengths of the requested I/O in the batch. The distance between requests is the amount of time needed by the disk to move from the end of one request to the beginning of the other, thus the amount of time needed to read the data itself, Transfer time , is not counted since it is independent of the order of the requests, only the order dependent Access time is computed.