This thesis has isolated the problem of deceptive idleness, and analyzed its impact on seek reducing and proportional-share schedulers. It has clearly demonstrated the benefits of using a non-work-conserving disk scheduler to solve deceptive idleness, and thus obtain significant improvements in throughput and adherence to quality of service objectives. This includes separate heuristics for seek reducing and proportional-share schedulers, to address their individual needs. This solution smoothly augments various workarounds like application and kernel level prefetching, and practically never causes degradation below a traditional system. It is easy to implement, and suited for incorporation into general-purpose operating systems.
Finally, this thesis has evaluated this solution for a variety of workloads. Microbenchmarks characterize the intrinsic properties of this solution, and real workloads test its effectiveness in more realistic circumstances. The Apache webserver delivers 56% and 16% more throughput for two configurations. The Andrew Benchmark runs faster by 8% (54% for the read-intensive phase). Variants of the TPC-B database benchmark exhibit improvements between 4% and 60%. Proportional-share schedulers become empowered to efficiently deliver application-desired proportions. All this, with almost no overhead.