Blaze MA. NFS tracing by passive network monitoring. Proceedings of the 1992 USENIX Winter Conference.USENIX Association: Berkeley, CA, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of client side caching than NFS. CAMPUS is an order of magnitude busier than any of the other systems, particularly in terms of the amount of data read and written. 4 Trace Analysis via NFS Passive NFS tracing is an old idea, and has been used in many trace studies and file system experiments [2, 3, 5, 7, 8]. The technique of passive tracing on a broadcast network evolved in parallel with broadcast net CAMPUS EECS CAMPUS EECS INS RES NT Sprite 9 1 11 30 9 1 11 30 10 21 10 27 10 21 10 27 Year of Trace 2001 2001 2001 2001 2000 2000 2000 1991 Days 91 91 7 7 31 31 31 8 Total ops (millions) 29.9 2.30 ....

....science department networks [7] but they received relatively little attention compared to kernel based trace studies or studies of other distributed file systems. This changed, however, as NFS became ubiquitous and better tracing tools and methodologies for analyzing NFS traces were devised [2, 8]. Passive NFS tracing is attractive from a research perspective because NFS is a portable and widespread protocol, used in a broad variety of real world contexts, and so its analysis is applicable to many interesting realworld problems. Unfortunately, we have found that even while computing has ....

[Article contains additional citation context not shown here]

Matthew A. Blaze. NFS Tracing by Passive Network Monitoring. In Proceedings of the USENIX Winter 1992.

....part) of the raw data that traverses the network and interpreting it at a later stage. Because of the capacity of networks (Ethernet has a raw capacity of 10Mb s) the sheer quantity of data does not easily allow for a raw trace of the network contents (difficult to store) Programs such as rpcspy [11, 12], built on the Sun and Digital network monitoring facilities mentioned above, do some processing and interpretation of incoming data to reduce the quantity of information ultimately recorded. Blaze, who constructed the rpcspy system of reference [11, 12] has used it to good effect for the ....

....to store) Programs such as rpcspy [11, 12] built on the Sun and Digital network monitoring facilities mentioned above, do some processing and interpretation of incoming data to reduce the quantity of information ultimately recorded. Blaze, who constructed the rpcspy system of reference [11, 12], has used it to good effect for the monitoring of a network based around a large file server. His research results have also been used in several other publications [39, 10, 15, 16, 13] Dahlin et al. 24] used the rpcspy tools to characterise file system load in a distributed system. The results ....

[Article contains additional citation context not shown here]

Blaze, M. NFS Tracing by Passive Network Monitoring. In USENIX Conference Proceedings, Winter 1992.

....an alternative technique for monitoring a computer s file system activities where the communications channel between a machine and its disk drives, and in particular between a diskless client and its disk server, is passively monitored. Blaze used this technique with his snooper rpcspy software [3], passively monitoring traffic between Network File System (NFS) 24] clients and servers, and predicting the operations the clients performed to cause those operations. Full kernel instrumentation is used commonly in system monitoring but, by definition, it involves the modification of the ....

Blaze, M. NFS Tracing by Passive Network Monitoring. In USENIX Conference Proceedings, Winter 1992.

....from a variety of research and instructional groups, and have diverse storage needs. This server is used on a daily basis in our department. Traces were collected using tcpdump on the network segment to which the server was attached. These packet observations were then fed into the nfstrace tool [2] which distilled the traces into individual NFS operations. Note that this tool captures client server interactions, but does not record operations satsified by a client s cache. Each operation log entry names a the target file and the host that issued the operation. Files are named in the raw ....

M. Blaze. NFS tracing by passive network monitoring. In Proceedings of the Winter 1992 USENIX Conference, pages 333--43, Berkeley, CA, USA, January 1992.

....performance analysis. Of more interest is DFSTrace, used by Mummert and Satyanarayanan [15] in the evaluation of the Coda le system, since they also replayed the traces using the timing information given by the trace. Instead of modifying the operating system kernel, Tourigny [17] and Blaze [6] exploited a remote le system architecture to obtain traces of le system activity by monitoring the interactions between clients and server. By contrast, we aim in this paper to capture the entire system call trace, and to use it to study the overall system performance by using it to reexecute ....

M. Blaze. NFS tracing by passive network monitoring. In USENIX Winter Conference, pages 333-334, 1992.

....file reference patterns. Of more interest is DFSTrace, used by Mummert and Satyanarayanan [11] in the evaluation of the Coda file system, since they also replayed the traces using the timing information given by the trace. Instead of modifying the operating system kernel, Tourigny [13] and Blaze [4] exploited a remote file system architecture to obtain traces of file system activity by monitoring the interactions between clients and server. This has the virtue of being entirely non intrusive, though includes only remote file accesses and also requires privileged access to the network. By ....

M. Blaze. NFS tracing by passive network monitoring. In USENIX Winter Conference, pp 333--334, 1992.

....used binaries on their local disks, such as files typically found in bin. Thus, references to these files do not appear in the traces. The fourth data set was collected by the creators of the experimental file system, xFS [10] to evaluate their system through simulation. Blaze s rpcspy program [17] was used to monitor network activity and generate traces for NFS clients on four Ethernets. During data collection, 4 of all network activity was dropped. Since close system calls do not appear in NFS network traffic, heuristics were used by Dahlin et al. in a postprocessing step) to add them. ....

M. Blaze, "NFS Tracing By Passive Network Monitoring," pp. 333-343 in Winter 1992 USENIX Conference Proceedings, San Francisco.

....without 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 ....

Matt Blaze. NFS Tracing by Passive Network Monitoring. In the Proceedings of the Winter USENIX Conference, pages 333--343. USENIX Association, Jan 1992.

No context found.

Blaze MA. NFS tracing by passive network monitoring. Proceedings of the 1992 USENIX Winter Conference.USENIX Association: Berkeley, CA, 1992.

No context found.

Matthew A. Blaze. NFS Tracing by Passive Network Monitoring. In Proceedings of the USENIX Winter 1992.

No context found.

Matthew A. Blaze. NFS Tracing by Passive Network Monitoring. In Proceedings of the USENIX Winter 1992.

No context found.

M. Blaze. NFS Tracing by Passive Network Monitoring. In Proceedings of the USENIX Winter Conference, January 1992.

No context found.

M. Blaze. NFS tracing by passive network monitoring. In Proceedings of the Winter 1992.

No context found.

Matthew A. Blaze, "NFS Tracing by Passive Network Monitoring," in Proceedings of the USENIX Winter 1992.

No context found.

M. Blaze. NFS tracing by passive network monitoring. In USENIX Conference Proceedings, pages 333--343, San Francisco, CA, Jan. 1992. USENIX.

No context found.

Matt Blaze. NFS Tracing by Passive Network Monitoring. In USENIX Winter Conference Proceedings, pages 333 -- 343. USENIX Association, January 1992.

No context found.

M. Blaze, "NFS Tracing By Passive Network Monitoring," pp. 333-343 in Winter 1992 USENIX Conference Proceedings, San Francisco.

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