Prediction of disk arm movements in anticipation of future requests

Algorithms for performing predictive movements of a disk arm in anticipation of future requests, aimed at reducing the disk access time, are described in this paper. The prediction is based on the past history of the requested disk tracks. A continuum of optimization algorithms has been derived, featuring different prediction performances and hence different complexities. Trace-driven simulations have shown that a relative seek-time reduction from about 11% to 23% for a workstation workload can be obtained using anticipatory movements.