Scheduling algorithms for modern disk drives
SIGMETRICS '94 Proceedings of the 1994 ACM SIGMETRICS conference on Measurement and modeling of computer systems
A hierarchial CPU scheduler for multimedia operating systems
OSDI '96 Proceedings of the second USENIX symposium on Operating systems design and implementation
Strategic directions in storage I/O issues in large-scale computing
ACM Computing Surveys (CSUR) - Special ACM 50th-anniversary issue: strategic directions in computing research
SOSP '01 Proceedings of the eighteenth ACM symposium on Operating systems principles
| Hi-index | 0.00 |
As the gap between processor and disk performance continues to grow in modern computing systems, so too does the need for improvements in disk performance management. In an effort to remove or reduce the performance bottleneck created by disk accesses, new approaches and algorithms for disk scheduling have been developed in recent years. While providing performance improvements, these approaches each have their own strengths and weaknesses that ultimately limit their applicability and usefulness across a wide variety of system workloads. This paper introduces a new method of disk access optimization which focuses primarily on dynamic scheduling algorithm selection and algorithm tuning. Disk activity is continuously collected in real-time and cached for later analysis to discover current system load patterns, while a scoring system is used to detect overall trends by system processes. Once analysis is complete, disk scheduling algorithms are automatically selected and/or tuned based on heuristics or criteria to ensure that the disk scheduling algorithm in use is well suited to the current workload of the system. Experimentation to date has been quite positive, demonstrating this approach has great potential for assisting in the optimization of system performance.