A description is given of Sprite, an experimental network operating system under development at the University California Berkeley. It part a larger research project, SPUR, for design and construction high-performance multiprocessor workstation with special hardware support Lisp applications. Sprite implements set kernel calls that provide sharing, flexibility, high performance to networked workstations. The discussion covers: application interface: basic structure; management file name...

Abstract The Sprite operating system allows executing processes to be moved between hosts at any time. We use this process migration mechanism offload work onto idle machines, and also evict migrated when workstations are reclaimed by their owners. Sprite's provides a high degree of transparency both for users. Idle machines identified, eviction is invoked, automatically daemon processes. On it takes up few hundred milliseconds on SPARCstation 1 perform remote exec, whereas evictions...

Data deduplication, an efficient approach to data reduction, has gained increasing attention and popularity in large-scale storage systems due the explosive growth of digital data. It eliminates redundant at file or subfile level identifies duplicate content by its cryptographically secure hash signature (i.e., collision-resistant fingerprint), which is shown be much more computationally than traditional compression approaches systems. In this paper, we first review background key features...

Caching in the World Wide Web currently follows a naive model, which assumes that resources are referenced many times between changes. The model also provides no way to update cache entry if resource does change, except by transferring resource's entire new value. Several previous papers have proposed updating entries only differences, or "delta," cached and current value.In this paper, we make use of dynamic traces full contents HTTP messages quantify potential benefits delta-encoded...

Mobile computers such as notebooks, subnotebooks, and palmtops require low weight, power consumption, good interactive performance. These requirements impose many challenges on architectures operating systems. This paper investigates three alternative storage devices for mobile computers: magnetic hard disks, flash memory disk emulators, cards.We have used hardware measurements trace-driven simulation to evaluate each of the their related design strategies. Hardware an HP OmniBook 300...

Much work on the performance of Web proxy caching has focused high-level metrics such as hit rates, but ignored low level details "cookies", aborted connections, and persistent connections between clients proxies well servers. These have a strong impact performance, particularly in heterogeneous bandwidth environments where network speeds are significantly different than We evaluate through detailed simulations latency effects environments. drive our with packet traces from two scenarios:...

Data-protection class workloads, including backup and long-term retention of data, have seen a strong industry shift from tape-based platforms to disk-based systems. But the latter are traditionally designed serve as primary storage there has been little published analysis characteristics workloads they relate design In this paper, we present comprehensive characterization by analyzing statistics content metadata collected large set EMC Data Domain systems in production use. This is both...

CPU speeds are improving at a dramatic rate, while disk not. This technology shift suggests that many engineering and office applications may become so I/O-limited they cannot benefit from further improvements. paper discusses several techniques for I/O performance, including caches, battery-backed-up cache logging. We then examine in particular detail an approach called log-structured file systems , where the system's only representation on is form of append-only log. Log-structured...

As data have been growing rapidly in centers, deduplication storage systems continuously face challenges providing the corresponding throughputs and capacities necessary to move backup within recovery window times. One approach is build a cluster system with multiple nodes. The goal achieve scalable throughput capacity using extremely high-throughput (e.g. 1.5 GB/s) nodes, minimal loss of compression ratio. key technical issue route intelligently at an appropriate granularity.We present...

Much work in the analysis of proxy caching has focused on high-level metrics such as hit rates, and approximated actual reference patterns by ignoring exceptional cases connection aborts. Several these low-level details have a strong impact performance, particularly heterogeneous bandwidth environments modem pools connected to faster networks. Trace-driven simulation pool large ISP suggests that "cookies" dramatically affect cachability resources; wasted due aborted connections can more than...

Modern storage systems orchestrate a group of disks to achieve their performance and reliability goals. Even though such are designed withstand the failure individual disks, multiple poses unique set challenges. We empirically investigate disk data from large number production systems, specifically focusing on impact failures RAID systems. Our covers about one million SATA six models for periods up 5 years. show how observed weaken protection provided by RAID. The count reallocated sectors...

Modern storage systems orchestrate a group of disks to achieve their performance and reliability goals. Even though such are designed withstand the failure individual disks, multiple poses unique set challenges. We empirically investigate disk data from large number production systems, specifically focusing on impact failures RAID systems. Our covers about one million SATA six models for periods up 5 years. show how observed weaken protection provided by RAID. The count reallocated sectors...

We examine the impact of using flash memory as a second-level file system buffer cache to reduce power consumption and access latency on mobile computer. use trace-driven simulation evaluate what we call FLASHCACHE. relate storage sub-system characteristics FLASHCACHE: its size, unit erasure, costs. find that FLASHCACHE can subsystem by 20-40% improve overall response time 30-70% when combined with an aggressive disk management policy. When more conservative policy, is reduced from 40-70%...