G. H. Kuenning, G. J. Popek, and P. L. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In Proceedings of the Summer 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....le systems, and or for le systems that provide built in support for concurrency [Stonebraker81, Stonebraker86] 3. Mobility: replicated and distributed le systems with disconnected and caching operations gure heavily in an environment where network latency and reliability is highly variable [Satyanarayanan90, Kistler91, Tait91, Tait92, Kistler93, Zadok93a, Kuenning94, Marsh94, Mummert95]. 4. Security: more secure le systems are sought, especially ones that securely export les over the network [Steiner88, Haynes92, Glover93, Takahashi95] An easy way to use encryption in le systems [Blaze93, Gutmann96, Boneh96] and the ability to provide special semantics via facilities such ....

G. Kuenning, G. J. Popek, and P. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. USENIX Summer 1994 Conference (To Appear) (Boston, U.S.), 1994.

....of our results. Finally, section 5 presents conclusions and discusses some issues resulting from our approach. 2 Related Work Recently lot of research has been directed at optimistic replication schemes. The problem has been well studied in the context of filesystems such as CODA [17] and FICUS [9]. Gray et al. in [7] present a system architecture and a replication model for mobile databases. BAYOU [4] provides another model for mobile data repositories. In BAYOU, information required for the reconciliation process is included in each update to the data repository. Our approach is closely ....

....non database systems which demonstrate such locality. Since fragments play such a crucial role in our architecture, we also need to provide efficient ways to define and create fragments. Furthermore, we would like to keep the so It is possible to work around this assumption using hoard profiles[9, 10]. Here the client need not explicitly specify which fragments to download. Instead, the client, say a sales manager on her way to a meeting in Singapore, can indirectly ask for data, as in, download everything related to Singapore . The database system can then use the hoard profile for Singapore ....

G. Kuenning, G. J. Popek and P. Reiher, An Analysis of Trace Data for Predictive File Caching in Mobile Computing, Proceedings of the USENIX Summer Conference, 1994, pages 291--303.

....the file system must provide optimistic replication of files. That is, multiple copies of files are allowed to exist and be updated independently, but attempts are made to detect and resolve update conflicts should any occur [11] Some studies have shown that update conflicts occur only rarely [2, 12, 7], and that optimistic replication is an acceptable strategy for a disconnected environment. 2.2. Weakly Connected Operation Weakly connected operation (also called partially connected [13] is similar to disconnected operation, but with a few important extensions. A key advantage is that cache ....

G. Kuenning, G. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. In Proc.

...., respectively. The system Cm i only needs to communicate with one server Sw j . The server Sw j will communicate with all the other Sw i if any updates occur. The mobile user is working on a set of working files. We use the average size of working file set and the average size of files given in [15]. Combining the data for productivity and programming environments, we use 30, 60 and 100 as the sizes of working files including data files and directories. This is controlled by the parameter wfsSize. The average length of data files is 27K. We assume that in a working file set 9 30 is ....

....by updtTime2. The write access rate of system Cm i is governed by writProb, which is assigned a value of 20 . The results of our simulation prove these updtTime1, updtTime2 and writProb are reasonable, since the write updating conflicts of files is about 0. 07 which is close to that reported in [15]. The CTO sharing semantics is chosen in our simulation. At the Sw j site, when a file directory is updated in a volume, the server Sw j will send a control message to all the Cm i that is communicating with this Sw j . If a system level volume is used, all updates against directories have to be ....

G.H. Kuenning, G.J. Popek and P.L. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In Proceedings of the 1994 Summer Usenix Conference.

....to a prior paper [33] However, our traffic generator is tailored to evaluate optimistic replication environments, and we will discuss this component in detail. Our simulation input traffic is based on a threemonth data trace of file access patterns, collected at Locus Computing Corporation [16]. This trace consists of the actual activities of software developers. The traced system did not use replication, but had large numbers of users and machines connected by a local area network, performing both local and remote file accesses. To a certain extent, the results presented depend on the ....

Kuenning GH, Popek GJ, Reiher PL. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. Proceedings of the 1994 Summer Usenix Conference, 1994.

....to prefetching based on data compression principles. Prefetching has also been studied in other systems environments. Prefetching in a parallel environment is studied in [KoE] More recently, the possibility of predictive prefetching in mobile computing has been studied by Kuenning et al. [KPR]. Cao et al. study integrated prefetching and caching strategies [CFK] in file systems assuming full knowledge of future requests; the concern in this case is to decide when to prefetch so as to minimize the total time consumed to satisfy a page request sequence. 2.2 Our Approach In Chapter 3, ....

....number of page faults NumFaults opt (oe) incurred by an optimal offline algorithm; i.e. NumFaults X (oe) c Delta NumFaults opt (oe) b: 3.1) In general, the cost of the algorithm replaces NumFaults in (3. 1) Competitive algorithms for cache replacement are well examined in the literature [BIR, FKL, IKP, KPR, McS, SlT]. It is unreasonable to expect algorithms to be competitive in this sense for prefetching. An optimal offline algorithm for prefetching would never fault, if it can prefetch at least one page every time. In order to be competitive, an online algorithm would have to be an almost perfect predictor ....

[Article contains additional citation context not shown here]

G. H. Kuenning, G. J. Popek, and P. L. Reiher, "An Analysis of Trace Data for Predictive File Caching in Mobile Computing," Proceedings of the 1994 Summer USENIX .

....in which items are requested from mass storage devices and enable 34 cache memories to minimize latency. They develop simulations based on a mass storage system modified to use TDAG. Using a synthetic user workload they show that such techniques show great promise. Kuenning, Popek, and Reiher [18] have done extensive work analyzing the behavior of file system requests for various mobile environments with the intent of developing a prefetching system that would predict needed files and cache them locally. Their work has concluded that such a predictive caching system has promise to be ....

Kuenning, G., Popek, G. J., and Reiher, P. An analysis of trace data for predictive file caching in mobile computing. In Proceedings of USENIX Summer Technical Conference (1994), USENIX, pp. 291--303.

....handle, H2 , for this newly created file. Therefore, RI needs to change the file handle information of the file in the local disk cache as well as the entries in the log file from H1 to H2 . 5. 4 Data Prefetcher (DP) Data prefetching is an effective technique for improving data access performance [18, 19]. In a mobile file system, it also increases the availability of data during the disconnection period by collecting data objects in the cache. Also, it reduces the latency during the weakly connected period by selectively decide which data need to be cached and which data need to be written back ....

Geoffrey H. Kuenning, Gerald J. Popek, Peter L. Reiher, An Analysis of Trace Data for Predictive File Caching in Mobile Computing, 1994 Summer USENIX Conference, April 1994

....and algorithms. 2.2 Prefetching The concept of prefetching has been used in a variety of environments including microprocessor designs, virtual memory paging, databases, and file read ahead. More recently, long term prefetching has been used in file systems to support disconnected operation [14, 15, 5]. Prefetching has also been used to improve parallel file access on MIMD architectures [4] One relatively straight forward method of prefetching is to have each application inform the operating system of its future requirements. This approach has been proposed by Patterson et al. 11] Using ....

Geoff Kuenning, Gerald J. Popek, and Peter Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In Proceedings of the 1994 Summer USENIX Conference, June 1994.

....section 6 presents conclusions and discusses new issues resulting from our approach to server organization for handling mobile clients. 2 Related Work As mentioned before, hoarding is a concept that has been successfully applied to file systems. Examples of such systems are CODA [26] and FICUS [17]. These systems allow clients to update replicas of files while disconnected, and reconcile updates on reconnection. CODA uses a single server to reconcile updates, whereas FICUS uses a peer to peer mechanism. Both use pessimistic conflict resolution to perform reconciliation of updates. In BAYOU ....

....data as in [15] Here data hoarded by the client is locked on the server until the client reintegrates its hoard 8 . Obviously this scheme is only practicable if multiple clients rarely attempt to access the same data. 7 It is possible to work around this assumption using hoard profiles[17]. Here the client need not explicitly specify which fragments to download. Instead, the client, say a sales manager on her way to a meeting in Singapore, can indirectly ask for data, as in, download everything related to Singapore . The database system can then use the hoard profile for Singapore ....

[Article contains additional citation context not shown here]

G. Kuenning, G. J. Popek and P. Reiher, An Analysis of Trace Data for Predictive File Caching in Mobile Computing, Proceedings of the USENIX Summer Conference, 1994, pages 291--303.

....100 bytes. Assume also that a user on an average accesses 200 new files (or files whose view entries are not in the database) every day. Note that the user may access a much larger number of files but the view entry of many of these files would be in the database because of file access locality [KPR94]. On this basis the view entry database for each entity grows by 20 KB per day. The rate at which the database can be truncated clearly depends on the frequency with which truncation is attempted. The average database size will become smaller as the frequency of truncation is increased. Other ....

Geoffrey H. Kuenning, Gerald J. Popek, and Peter Reiher. "An Analysis of Trace Data for Predictive File Caching in Mobile Computing." In USENIX Conference Proceedings, pp. 291--306. USENIX, June 1994.

....timing information, is an inaccurate measure and over estimates the potential performance improvements. Third, the traces must monitor all file system activity including each read and write operation, not just open events. Public traces such as the Sprite traces [BHK 91] and others [Bla92, KPR94, DMW 94] do not record all file system activity or include the timing information we needed. 5 Recently, some new file tracing tools have become available that did not exist when we began our work[MS94, ES94] 4.1.2 Information Gathered Because the trace information we needed was not ....

.... in a variety of environments including microprocessor designs, virtual memory paging, databases, and file read ahead [AK96, Mow94, PZ91, Smi78, Smi85] In the context of mobile computing, long term (batch) file prefetching has been used to support disconnected operation [Sat90, SK93, Kue94, KPR94, ABC89, HH93] File system prefetching has long been used to improve intrafile accesses (accesses within a file) Smith [Smi78] and Korner [Kor90] examined the problem of predicting future file block accesses within the current file. Such models can significantly improve access to large files ....

[Article contains additional citation context not shown here]

G.H. Kuenning, G.J. Popek, and P.L. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In the Proceedings of the 1994 Summer USENIX Conference, pages 291--303, June 1994.

No context found.

G. H. Kuenning, G. J. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. In USENIX Conference Proceedings, pages 291--306. USENIX, June 1994.

....writers within the large quantity of replicas. While the hot spots may change over time, it is rare to see a widely replicated object that is consistently updated by everyone. Simulation data by Wang supports the hot spot notion. In studies of productivity environment data taken at Locus Computing [4], Wang found that replicated objects tend to have only a small set of common writers at any one time [15] In short, while we must provide the ability for everyone to generate updates, it seems rare that a significant percentage of the replicas are simultaneously trying to do an assertion we ....

G. H. Kuenning, G. J. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. In USENIX Conference Proceedings, pages 291--306. USENIX, June 1994.

No context found.

Geoffrey H. Kuenning, Gerald J. Popek, and Peter Reiher, `An analysis of trace data for predictive file caching in mobile computing', USENIX Conference Proceedings. USENIX, June 1994, pp. 291--306.

No context found.

Kuenning, G. H., Popek, G. J., and Reiher, P. L., "An Analysis of Trace Data for Predictive File Caching in Mobile Computing," in Proceedings of the 1994 Summer Usenix Conference, 1994.

No context found.

G. H. Kuenning, G. J. Popek, and P. L. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In Proceedings of the Summer 1994.

No context found.

G. Kuenning, G. J. Popek, and P. Reiher, "An analysis of trace data for predictive file caching in mobile computing," in Proceedings of the USENIX Summer Technical Conference, pp. 291-- 303, USENIX, 1994.

No context found.

G. H. Kuenning, G. J. Popek, and P. L. Reiher, "An analysis of trace data for predictive file caching in mobile computing", USENIX Summer Conf., 1994, pp. 291--303.

No context found.

G. Kuenning, G. J. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. Proceedings of the USENIX Summer Conference, pages 291--303, 1994.

No context found.

G. Kuenning, G. J. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. Proceedings of the USENIX SummerConferenc,pages 291--303, 1994.

No context found.

G. Kuenning, G. J. Popek, and P. Reiher. An analysis of trace data for predictive file caching in mobile computing. Proceedings of the USENIX Summer Conferenc, pages 291--303, 1994.

No context found.

G. H. Kuenning, G. Popek and P. Reiher, "An Analysis of Trace Data for Predictive File Caching in Mobile Computing". 1994 Summer Usenix Conference, 1994.

No context found.

Geoffrey H. Kuenning, Gerald J. Popek, and Peter L. Reiher. An Analysis of Trace Data for Predictive File Caching in Mobile Computing. In USENIX Summer Conference Proceedings. USENIX Association, June 1994.

No context found.

Kuenning GH, Popek GJ and Reiher P, "An analysis of trace data for predictive file caching in mobile computing", USENIX Conference Proceedings, pages 291-306. USENIX, June 1994.

First 50 documents

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