On Improving the Transient Performance of CSCAN Schedulers

Abstract

In order to achieve high stream throughput, disk overheads can be minimized by seek-reducing disk scheduling techniques, such as SCAN and CSCAN, in which services to streams are ordered according to disk cylinder positions. Unfortunately, initial service to new streams cannot be immediate because this will introduce extra seek overhead into a service period. To avoid violating real-time guarantees, static seek-reducing VOD schedulers may need to delayed starting of new streams by up to two service periods [9, 10]. Consequently, the start-up latency of existing static seek-reducing schedulers is long compared to round-robin and deadline-driven approaches.