P. Cao, E. W. Felten, A. Karlin and K. Li, "Implementation and Performance of Integrated Application-Controlled Caching, Prefetching and Disk Scheduling", ACM Transaction on Computer Systems, November 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....controller replies to the request with several blocks. 3 Related Work There have been several works on improving disk I O performance. Most of them discuss techniques that are external to the disk drive. Typical examples are techniques for improving the prefetching and caching of disk blocks [21, 6, 27], scheduling requests [3, 15] and for optimizing disk writes [13, 23, 32] Our techniques deal with read ahead (a form of prefetching) and caching of disk data, but their focus is on the disk controller cache. There are several important differences between the controller cache and other caches, ....

....sophisticated cache management techniques. Furthermore, FOR and HDC require only minimal programmability to accommodate different file systems and workloads. Finally, more loosely related are works on main memory caching for servers, e.g. 19, 2] and application controlled main memory caching [6]. Our work is complementary to the server caching studies in that we focus on the disk controller cache, which exhibits significantly different characteristics than a main memory cache, as mentioned above. Our work differs from application controlled caching as well. In this approach, applications ....

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311--343, November 1996.

....optimal performance. Rajasekaran [22] gives another asymptotically suboptimal deterministic parallel disk sorting algorithm that runs in three passes for not too large inputs. Prefetch bu#ers for disk load balancing and overlapping of I O and computation has been intensively studied [21, 7, 3, 14, 13, 12]. But we have not seen results that guarantee overlapping of I O and computation during parallel disks merging of arbitrary runs. There are many good practical implementations of sorting (e.g. 19, 1, 30, 20] that address parallel disks, overlapping of I O and computation, and low internal ....

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated application-controlled file caching, prefetching and disk scheduling. ACM Transactions on Computer Systems, 14(4):311--343, Nov. 1996.

....access times have become increasingly important. File prefetching has been widely researched in order to address the problem of slow disks, and particularly disk access latency. Much of the research has focused on how to best predict future accesses [2] 5] 13] 10] 12] though some researchers [4][15] have examined the question of when to perform prefetches, assuming knowledge of future accesses. Patterson introduced cost benefit analysis for file prefetching, which seeks a good balance between caching recently used blocks and prefetching blocks which are predicted for future use (Figure ....

Cao, Pei, Edward W. Felten, Anna R. Karlin and Kai Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching and disk scheduling. A CM Transactions on Computer Systems, 14(4):311-343, November 1996.

....of inappropriate or incomplete file system heuristics is for the application using the files to supply hints to the file system about their anticipated access patterns. These hints can be extremely successful, especially when combined with the techniques of prefetching and selective caching [3, 12]. The drawback with this approach is that it requires that applications be modified to provide hints. There has been work in having the application compiler automatically generate hints, but success in this area has been largely confined to scientific workloads with highly regular access patterns ....

Pei Cao, Edward W. Felten, Anna R. Kar- lin, and Kai Li. Implementation and performance of integrated application-controlled file caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14(4):311-343, 1996.

....for the longest time. This policy exploits temporal locality in data accesses (the observation that, in general, data that has been accessed recently is likely to be accessed in the near future) and, while a mismatch for some access patterns, is known to be fairly effective in general. Prior work [43, 3, 4, 24, 62] has proposed a variety of algorithms for leveraging more explicit information about what data will be accessed in the future. Because the system implemented and evaluated in this dissertation takes advantage of TIP2 [43] the rest of this section focuses on TIP2, hereafter referred to simply as ....

Pei Cao, Edward W. Felten, Anna R. Karlin, and Kai Li. Implementation and performance of integrated application-controlled file caching, prefetching and disk scheduling. ACM Transaction on Computer Systems (TOCS), 14(4):311--343, 1996.

....to track replacement costs. Second, broadcast disks assume an infrequently changing broadcast schedule, whereas storage aware caching must react to frequent changes in workload and device performance. Recently, researchers have studied allocation of pages between different classes in prefetching [9, 24, 35], compiler controlled memory management [14] and resizeable file buffer caches [22] In prefetching, page allocation occurs between applications [9] or hinted and unhinted I O references [24, 35] For compiler controlled memory management, the compiler provides application memory usage ....

....changes in workload and device performance. Recently, researchers have studied allocation of pages between different classes in prefetching [9, 24, 35] compiler controlled memory management [14] and resizeable file buffer caches [22] In prefetching, page allocation occurs between applications [9] or hinted and unhinted I O references [24, 35] For compiler controlled memory management, the compiler provides application memory usage information to operating system global replacement policies using hints, reintegrating elements of local page replacement into global page replacement. ....

[Article contains additional citation context not shown here]

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching, and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311--343, November 1996.

....RAID scrubbing, virus detection, and backup. Another set of examples is the many defragmentation [15, 29] and replication [18, 31] techniques that have been developed to improve the performance of future accesses. A third set of examples is anticipatory disk activities such as prefetching [7, 11, 13, 19, 27] and prewriting [2, 4, 8, 10] Using simulation, our previous work explored two specific uses of freeblock scheduling. One set of experiments showed that cleaning in a log structured file system [22] can be done for free even when there is no truly idle time, resulting in up to a 300 increase in ....

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14(4):311 343, November 1996.

....show that LRU (Least Recently Used) is the replacement algorithm of choice when caching NFS requests on an NFS server. Within the last few years other replacement algorithms have been introduced, including a compression based algorithm in Phalke[2] and an application controlled algorithm in Cao[3][4]. Phalke shows that compression of the IRG (Inter Reference Gap) can be used to predict replacement and improve perfor 129 mance. Cao allows application some control over placement, which can dramatically increase performance, but requires users to have inside knowledge of the cache ....

....this allows us to decide if replacement cost can be used to evaluate cache performance by means of overall usage of system resources. The flexibility introduced by using a weighting function can be used to include even application induced replacement criteria. This will allow the results of Cao[3][4] to be integrated. To further improve cache performance other methods, such as prefetching (see Bennett[13] should be considered as well. References [1] B. Reed, D.D. Long, Analysis of caching Algorithms for distributed file systems, Operating Systems Review, Vol. 30, No. 3, 1996 [2] V. ....

P. Cao, E. Felden, A. Karlin, K. Li, Implementation and performance of integrated applicationcontrolled caching, prefetching and disk scheduling, Proceedings of the ACM SIGMETRICS conference on measurement and modeling of computer systems, 1995

....all the blocks accessed by the application in advance. 3 Related Work There have been several works on improving disk I O performance. Most of them discuss techniques that are external to the disk drive. Typical examples are techniques for improving the prefetching and caching of disk blocks [16, 4, 24], scheduling requests [2, 11] and for optimizing disk writes [10, 18, 28] Our techniques deal with read ahead (a form of prefetching) and caching of disk data, but their focus is on the disk controller cache. There are several important differences between the controller cache and other caches, ....

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311--343, November 1996.

....(as in FFS s cylinder groups) are no longer operational. Data caching efforts have focused on aggressive write back and prefetching to hide lengthy disk access times. In particular, many schemes have been proposed for identifying access patterns [28] allowing application and compiler hinting [6, 36, 5] and even speculatively executing programs to get hints [10] There are also application interfaces, such as dynamic sets [51] and disk directed I O [27] that allow groups of requests to be specified collectively. Similarly, schemes like soft updates [17] and RIO [11] allow aggressive write back ....

....in log structured file systems as a concrete example of such use of free bandwidth. Prefetching and prewriting. Another use of free bandwidth is for anticipatory disk activities such as prefetching and prewriting. Prefetching is well understood to offer significant performance enhancements [36, 6, 21, 28, 50]. Free bandwidth prefetching should increase performance further by avoiding interference with foreground requests and by minimizing the opportunity cost of aggressive predictions. As one example, the sequence shown in Figure 3(b) shows one way that the prefetching common in disk firmware could be ....

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14 (4): 311343, November 1996.

.... over multiple requesting applications, and reduce processing overheads by coalescing distinct accesses into a single larger access [38, 55] For example, algorithms have been developed for eiticient use of device parallelism and cache resources, given suiticient access pattern information or hints [7, 36, 54]. Useful extensions would be to support multiple active streams with different information sources (type, access pattern, service parameter) different accuracies of advance knowledge, and different service requirements. Storage systems for continuous media, and the associated bandwidth and ....

CAO, P., FELTEN, E. W., KARLN, A., AND LL K. Implementation and performance of integrated application-controlled caching, prefetching and disk scheduling. ACM Transactions on Computer Systems (to appear). An earlier version available as Technical Report CS-TR-94-493, Princeton University.

No context found.

P. Cao, E. W. Felten, A. Karlin and K. Li, "Implementation and Performance of Integrated Application-Controlled Caching, Prefetching and Disk Scheduling", ACM Transaction on Computer Systems, November 1996.

No context found.

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching and disk scheduling. ACM Transactions on Computer Systems, 14(4):311-- 343, 1996.

No context found.

P. Cao and E. Felten, Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching, and Disk Scheduling, ACM Transactions on Computer Systems, vol. 14, No. 4, 1996

No context found.

Pei Cao and Edward W. Felten. Implementation and Performance of Integrated AppplicationControlled File Caching, Prefetching, and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311-343, November 1996.

No context found.

P. Cao, E. Felten, A. Karlin, and K. Li. Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching, and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311--343, Nov. 1996.

No context found.

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated application-controlled le caching, prefetching and disk scheduling. ACM Transactions on Computer Systems, 14(4):311-343, 1996.

No context found.

Pei Cao, Edward W. Felten, Anna R. Karlin, and Kai Li. Implementation and Performance of Integrated Application-Controlled File Caching, Prefetching, and Disk Scheduling. ACM Transactions on Computer Systems, 14(4):311--343, 1996.

No context found.

Pei Cao, Edward W. Felten, Anna R. Karlin, and Kai Li. Implementation and performance of integrated application-controlled file caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14(4):311--343, November 1996.

No context found.

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14(4):311--343, November 1996.

No context found.

P. Cao, E. W. Felten, A. Karlin, and K. Li, "Implementation and performance of integrated application-controlled file caching, prefetching, and disk scheduling", ACM Transactions on Computer System, 14(4):311-343, November 1996.

No context found.

P. Cao, E. W. Felten, A. Karlin, and K. Li, Implementation and performance of integrated application-controlled file caching, prefetching, and disk scheduling, ACM Transactions on Computer System, 14(4):311-343, November 1996.

No context found.

Pei Cao, Edward W. Felten, Anna R. Karlin, and Kai Li, "Implementation and performance of integrated application-controlled file caching, prefetching, and disk scheduling," ACM Transactions on Computer Systems, volume 14, number 4, pages 311--343, 1996.

No context found.

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled le caching, prefetching, and disk scheduling. ACM Transactions on Computer Systems, 14(4):311{ 343, November 1996.

No context found.

P. Cao, E. W. Felten, A. R. Karlin, and K. Li. Implementation and performance of integrated applicationcontrolled file caching, prefetching and disk scheduling. ACM Transactions on Computer Systems, 14(4):311-- 343, 1996.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC