C. Staelin, H. Garcia-Molina, "Smart File Systems," Proc. of USENIX Winter '91, pp 45-51.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and prefetching files in anticipation of future requests. An algorithm for selecting files to prefetch based on second order correlations despite the presence of noise is presented. Cluster based disk reorganization was unable to achieve any benefit over traditional block and cylinder shuffling [21, 24, 23, 1]. Our prefetching systems were able to predict files to prefetch with over 50 accuracy. 1 Introduction Microprocessor performance has increased 50 per year since the mid 1980 s [18] however, improvements in total system performance have been unable to keep pace. One key limiting factor to ....

....correlations and is effective even in the pres1 ence of noise in the access stream. High level simulations of the Clump system return mixed results. The cluster based reorganization of the file system produced no improvement in seek time compared with techniques presented in the literature [1, 21, 23, 24]. The prefetching technique has shown to be very effective at predicting future file requests, although our current implementation is unable to convert a high prediction rate into improved system performance. The remainder of this paper is organized as follows: section 2 describes related work, ....

[Article contains additional citation context not shown here]

C. Staelin and H. Garcia-Molina. Smart filesystems. Proceedings of the Usenix Winter

....access frequencies. However, if it occupies holes that are frequently reused, shuffling the holes can still produce benefit. A variant on physical shuffling is to perform logical shuffling, for example, at the level of entire files. This is the approach adopted in the iPcress file system [Staelin91]. An obvious advantage of this technique is that it can more easily keep associated data together; disadvantages are assumptions that an entire file has a common access rate, that intra file data associations are more important than inter file associations, the variable sized nature of the objects ....

Carl Staelin and Hector Garcia-Molina. "Smart filesystems." Proceedings of Winter 1991 USENIX (Dallas, TX), pages 45--51, 21--25 January 1991.

....the above two techniques are best applied in the disk itself because they rely on fast access to rotation position information, the next idea could be applied in the file system as well possibly with even better results. Indeed, some similar work has been done in this area on wholefile layout [Staelin91]. The presence of the Loge indirection table gives us additional benefits beyond the fast writes: we can now consider optimizing the layout of the disks on a per block basis for better read performance, for example, by shuffling the most actively read data to the center of the disk. An ....

Carl Staelin and Hector Garcia-Molina. Smart filesystems. In Proceedings of the Winter 1991 USENIX (Dallas, TX), pp. 45--51, 21--25 January 1991.

....the leastaccessed data has been placed at the edge of the disk. More recently, similar results have been shown for optical storage media [Ford 91] In practice, data references are not drawn from a fixed distribution, nor are they independent. Although references are highly skewed [Floyd 89, Staelin 91, Vongsath 90, Ouster 85] request distributions change over time, and they are generally not known in advance. Nevertheless, variations of the organ pipe heuristic seem to work well in practice. Recently, several papers have proposed adaptive applications of data clustering based on this idea. ....

....Recent work in the DataMesh project [Ruemmler 91] considered rearrangement of cylinders and blocks, with mixed results. Their conclusion that block shuffling generally outperforms cylinder shuffling corroborates one of our own. A similar approach is employed in the experimental iPcress file system[Staelin 91] which monitors access to files and moves files with high temperatures (frequency of access divided by file size) to the center of the disk. Our technique differs from each of the techniques mentioned above in at least one of the following respects. Granularity Our technique moves blocks ....

[Article contains additional citation context not shown here]

Staelin, Carl, Hector Garcia-Molina, "Smart Filesystems," Proceedings of the Winter 1991 USENIX Conference, Dallas, TX, 1991.

....work in the DataMesh project [Ruemmler 91] considered rearrangement of cylinders and blocks, with mixed results. Their conclusion that block shuffling generally outperforms cylinder shuffling corroborates one of our own. A similar approach is employed in the experimental iPcress file system [Staelin 91] which monitors access to files and moves files with high temperatures (frequency of access divided by file size) to the center of the disk. The technique presented here differs from these in several ways: ffl We move blocks, rather than cylinders or files. ffl We rearrange only a small ....

....simulation studies driven by traces from a VMS system. We conclude with a summary of our results and possible extensions to our study. 2 Adaptive Rearrangement Overview The idea of block rearrangement is motivated by the fact that access to data stored on disks is highly skewed[Floyd 89, Staelin 91, Ruemmler 91] Of the thousands of data blocks stored on a disk, a small fraction of them absorbs most of the requests. If the hot (frequently accessed) blocks are spread over the surface of the disk, distant from each other, long seek delays may result. Hot blocks can be clustered to reduce ....

Staelin, Carl, Hector Garcia-Molina, "Smart Filesystems," Proceedings of the Winter 1991 USENIX Conference, Dallas, TX, 1991.

....out. Violates NFS semantics. # Consider track remapping. Not on the controller unless the controller is physically part of the drive. Do it in the driver (change the fs magic #) or the drive itself. Let Fuller do it it s his idea. The motivation and justification for this idea is explained in [Staelin]. 10. ....

Carl Staelin, "Smart file systems," Proceedings of the Usenix Technical Conference, Winter '91. Draft of 14:51:21 2/7/91 of @(#)nvram 1.11

....at 30 minutes with a worst case of approximately one hour. This would clearly be unacceptable in a system which is in constant use. However, the emphasis of the study was the design of an intelligent disk controller. In the future, smart , automatically adaptive disks may become available. Staelin and Garcia Molina (1991) describe the implementation of a smart file system, iPcress. IPcress gathers statistical information on each file and assigns each one a hotness. Files are ranked according to their hotness and rearranged CHAPTER 2. PERSISTENT STORE DESIGN 34 to place the hottest files at the centre of ....

....read has been hidden beneath the computation and the overall time for the computation has been reduced. The UNIX operating system prefetches blocks of a file when it detects sequential access is occurring (Stonebraker et al. 1983) In iPcress, small files are prefetched into the cache on open (Staelin Garcia Molina, 1991). Stonebraker (1981) states that the INGRES database management system can nearly always predict which block it will access next. This is not always the next logical block and therefore UNIX would not be able to prefetch this block. Patterson et al. 1993) argue that prefetching can be extended ....

Staelin, C., & Garcia-Molina, H. 1991. Smart Filesystems. Pages 45--51 of: USENIX Dallas Conference Proceedings.

....general performance in various ways. Sprite LFS clusters I O requests, uses large caches and, like VIFS, uses asynchronous creates and deletes [5, 11] to achieve better file system performance. Others use heuristics to tune caching and storage allocation algorithms for current usage patterns [13]. VIFS is related to these systems because it attempts to group blocks in a file close together and the allocation algorithm changes slightly as the file size grows. It differs from these file systems in that it achieves most of its performance from the layout scheme rather than fancy buffering ....

Carl Staelin and Hector Garcia-Molina. Smart filesystems. In USENIX Conference Proceedings, pages 45--51, Dallas, Texas, January 1991.

....objects, each optimized for a particular type of read access. A storage object can implement multiple policies and keep statistics (possibly stored in the meta data) to determine which policy is most effective for that data. A similar approach is used by Staelin et al. for reorganizing data on disk [25], and by Kotz to prefetch data from disk [15] Finally, an application can derive a new storage object from a file system storage object at run time. The ELFS file system [10, 11] uses such an approach both to invoke application specific policies and to close the semantic gap between the ....

C. Staelin and H. Garcia-Molina. Smart Filesystems. In Winter USENIX, 1991.

....to improve the flexibility of file systems in order to make it easier to add new features such as compression and encryption. Two examples that use an object oriented approach are the Spring file system [61] and the Choices file system [15] The iPcress file system by Staelin and Garcia Molina [122] also uses (in a more limited fashion) an object oriented approach, where each file is a file object that maintains statistics on the usage of the file to determine how data should be cached and how the file should be stored on disk. Stackable file systems are another approach for achieving the ....

C. Staelin and H. Garcia-Molina. Smart filesystems. In Winter USENIX, pages 45--51, 1991.

....clustering algorithms. Adding the clustering algorithm to iPcress improved disk performance 10 30 . Some of the material presented in this dissertation has been published previously. Material from Chapter 3 appeared in Reference [110] Chapters 4 and 5 contain material from References [112] and [113], and Chapter 7 contains material from Reference [111] Chapter 2 Prior Work In order to design a high performance file system, we must first understand how existing file systems are used, so we analyze file access patterns in existing systems. We describe several file systems and categorize ....

C. Staelin and H. Garcia-Molina. Smart filesystems. In Proceedings Winter 1991 USENIX, pages 47--54, Dallas, Jan. 1991.

No context found.

C. Staelin, H. Garcia-Molina, "Smart File Systems," Proc. of USENIX Winter '91, pp 45-51.

No context found.

C. Staelin and H. Garcia-Mollina. Smart Filesystems. In Proceedings of the USENIX Winter Technical Conference (USENIX Winter '91), pages 45--51, Dallas, Texas, January 1991.

No context found.

C. Staelin and H. Garcia-Molina, Smart filesystems, In USENIX Winter 1991.

No context found.

C. Staelin and H. Garcia-Molina. Smart filesystems. In Proceedings of the Winter 1991.

No context found.

C. Staelin and H. Garcia-Molina. Smart filesystems. In Proceedings of the Winter 1991.

No context found.

C. Staelin and H. Garcia-Molina. Smart Filesystems. In Proceedings of the 1991.

No context found.

C. Staelin and H. Garcia-Molina, `Smart Filesystems', Proceedings of the 1991 Winter USENIX, pp. 45-51, Dallas, TX, 1991.

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