``DULO: an effective buffer cache management scheme to exploit both temporal and spatial localities" Song Jiang, Xiaoning Ding, Feng Chen, Enhua Tan, and Xiaodong Zhang Proceedings of the 4th USENIX Conference on File and Storage Technologies (FAST'05), San Francisco, CA, December 14-16, 2005. Abstract A hard disk drive has a non-uniform access performance that is heavily affected by the locations of requested blocks and disk head during accesses. Sequentiality of requested blocks on disks, or their spatial locality, is critical to the performance of disks, where accesses to a sequentially placed disk blocks can have an order of magnitude higher bandwidth than that of the accesses to the randomly placed blocks. While I/O scheduling and prefetching can effectively exploit spatial locality and dramatically improve disk bandwidth for workloads with dominant sequential accesses, their ability to deal with workloads mixed with sequential and random data accesses, such as those in Web services, databases, and scientific computing applications, is very limited. Since spatial locality of cached blocks is largely ignored and only temporal locality is considered in system buffer cache management, disk performance for workloads without dominant sequential accesses can be seriously degraded. To address this problem, we propose a scheme called DULO (DUal LOcality), which exploits both temporal and spatial localities in buffer cache management. Leveraging the filtering effect of the buffer cache, DULO can influence the I/O request stream by making the requests passed to disk more sequential, significantly increasing the effectiveness of I/O scheduling and prefetching for disk performance improvements. DULO has been intensively evaluated by both trace-driven simulations and its implementation in the most recent Linux kernel 2.6.11. In simulations and system measurements, various application workloads have been tested, including Web Server, TPC benchmarks, and scientific application programs. Our experiments show it can significantly increase system throughput and reduce program execution times.Back to the Publication Page.
Back to the HPCS Main Page at the Ohio State University.