Using dynamic sets to reduce the aggregate latency of data access

摘要: Many users of large distributed systems are plagued by high latency when accessing remote data. Latency is particularly problematic for the critical application search and retrieval, which tends to access many objects may suffer a long wait each object accessed. Existing techniques like caching, inferential prefetching, explicit prefetching not suited search, ineffective at reducing applications, or greatly increase complexity programming model. This dissertation shows that extending file system interface support new abstraction called dynamic sets can address problem without incurring penalties other techniques. A set lightweight transitory collection with well-defined semantics. An creates on-demand hold it wishes process. Adding system's results in two benefits. First, creation discloses application's interest set's members system. allows reduce aggregate I/O through reordering requests. Second, provide direct manipulating objects. Thus improve performance functionality unduly increasing model. This describes design abstraction, an implementation adds 4.3BSD interface, evaluation implementation. The several including Unix tools WWW browser, Coda, NFS, WWW, local With little effort one modify applications use extend allow systems. substantially on wide area For example, replaying traces real searching over order magnitude across range factors.