Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. ACM SIGMETRICS Conference on Measurement and Modeling of Computer Systems (Madison, WI, 21--26 June 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the migration executor moves data across devices to complete the migration plan. Excessive migrationpenalties are unacceptable for applications with QoS requirements. Due to the uncertainties in storage systems and the fact that existing workload models allow fluctuations on arbitrary time scales [35], it is difficult to a priori predict the correct speed of data migration that will not cause QoS violations and not too pessimistic at the same time. In this thesis, we present Aqueduct, a feedback control architecture for the migration executor. 6.2.1. Migration Planner The migration planner ....

S. Gribble, G. Manku, E. Roselli and E. Brewer, "Self-similarity in File Systems", SIGMETRICS'98, April 1998.

....Performance Council) test on an Oracle database system, under various configurations, to demonstrate the effectiveness of our technique. General Terms Algorithms, Measurement, Performance. Keywords Scalability, Data Mining, Regression. 1. INTRODUCTION Solving system performance problems [3] [9] [13] 14] 15] is becoming increasingly crucial due to the explosive growth in the scale and complexity of information systems. Many organizations provide online transaction processing and other services, so good performance of their computing systems is vital to their existence. When performance ....

....Their output is usually a time series. Determining which system resources to examine is often a lengthy process of educated guesswork, where many problems can run undetected. However, few attempts have been made to capture the similarities and differences under various system configurations. In [9], the authors show that self similar ( 4] 11] behaviors exist in highlevel file system events. However, in the paper, they only focus on the file system and only very few attributes are examined. 3] 15] 13] 14] 10] 17] and numerous other papers also deal with performance issues, but are ....

Steven D. Gribble, Gunneet Singh Manku, Drew Roselli, and Eric A. Brewer, Timothy J. Gibson and Ethan L. Miller, "Self-Similarity in File Systems", SIGMETRICS '98 Madison, Wl, USA.

....Whereas proxies have been shown to be effective in reducing the mean of this distribution, their effectiveness in reducing the tail of the distribution has not been quantified. Impact of burstiness on locality: Web workloads exhibit burstiness at multiple time scales (see Figures 1(b) and (c) [11, 16]. Periods of intense bursts in such workloads govern the tail of the load distribution. If the locality exhibited by the workload reduces during intense bursts, then the tail of the distribution will be relatively unaffected even in the presence of a proxy cache (since the reduced locality will ....

....we examine (a) whether proxies can help smooth out the burstiness in web workloads, and (b) the relationship between locality and the burstiness of web workloads. 3.3. 1 Effect of Caching on Burstiness Past studies have shown that web workloads exhibit burstiness at multiple time scales [11, 16]; the workloads employed in our study are no different (see Figures 1(b) and (c) In this section, we determine whether a proxy cache can help smooth out the burstiness in web workloads, or whether the resulting workload becomes more bursty. A better understanding of these issues will help ....

[Article contains additional citation context not shown here]

S. D. Gribble, G. Manku, D. Roselli, E. Brewer, T. Gibson, and E. Miller. Self-Similarity in File Systems. In Proceedings of ACM SIGMETRICS '98, Madison, WI, June 1998.

....on the fly and learn the traffic characteristics in real time. 3 Background: Self Similarity After the initial discovery that Ethernet traffic is self similar [9] a high degree of self similarity has been observed in many other types of traffic (e.g. TCP [11] video [5] web [3] file system [8], and disk I O [6] traffic) In this section we give a brief overview of self similar processes. Informally, self similarity means invariance with respect to scaling across all time scales. Invariance may mean exact 2 identity, in which case we speak of deterministic self similarity. However, ....

....of 0.7. Both the algorithms show linear scalability. 6 Conclusions Our proposed method is very general in the sense that such self similar, bursty time sequences arise very often in real world data. This was recently observed in numerous settings, like TCP [11] video [5] web [3] file system [8], and disk I O [6] traffic. 10 The main contribution of this work is the introduction of the b model as an effective tool for finding and characterizing patterns in real, bursty time sequences. The model is extremely compact, as it effectively needs only one parameter, the bias b. Additional ....

Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. In SIGMETRICS'98, 1998.

....on the fly and learn the traffic characteristics in real time. 3 Background: Self Similarity After the initial discovery that Ethernet traffic is self similar [9] a high degree of self similarity has been observed in many other types of traffic (e.g. TCP [11] video [5] web [3] file system [8], and disk I O [6] traffic) In this section we give a brief overview of self similar processes. Informally, self similarity means invariance with respect to scaling across all time scales. Invariance may mean exact identity, in which case we speak of deterministic self similarity. However, it ....

....distribution for lbl network traces 6 Discussion and Conclusion Our proposed method is very general in the sense that such self similar, bursty time sequences arise very often in real world data. This was recently observed in numerous settings, like TCP [11] video [5] web [3] file system [8], and disk I O [6] traffic. The main contribution of this work is the introduction of the b model as an effective tool for finding and characterizing patterns in real, bursty time sequences. The model is extremely compact, as it effectively needs only one parameter, the bias b. Additional ....

Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. In SIGMETRICS'98, 1998.

....Although this is a simpli cation, it has been shown to closely approximate real measurements over all but the shortest time intervals [17] Each Poisson source has an average rate of 100 kbps. 2. FTP of large les: Since studies of Ethernet and LAN trac have shown the trac to be self similar [7, 12], a good test of the proposed algorithm would be to see the e ect on delay jitter of the video sources in the presence of self similar cross trac. In this connection, note that FTP of a large le between two hosts on the same Ethernet segment represents the worst case from the point of view of ....

S. Gribble, G. Manku, D.Roselli, E. Brewer, T. Gibson, and E. Miller. Self-Similarity in File Systems. In Proceedings of the ACM SIGMETRICS, pages 141-150, June 1998.

....0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 Log10(d) Figure 3. Pox Plot of CPU Availability using Unix Load average from thing1 (left) and thing2 (right) fic [20] World Wide Web traffic [8] network protocol performance [22] transmitted video traffic [3] and networked file systems [17] all point to self similarity as an inherent property of modern distributed systems. Of particular interest is the work by Dinda and O Halloran [10] in which the authors rigorously analyze Unix load average data from a large number of computational settings. The focus of their analysis is on the ....

....The pervasiveness of these observations across our set of experiments supports the previous work of Dinda and O Halloran. We, therefore, surmise that the CPU availability exhibits longrange autocorrelation and is either self similar (as noted in [10] or short term self similar as described in [17]. 3.2 Longer term prediction of CPU availability Despite the long range autocorrelation present in CPU availability measurements, the data in Tables 2 and 3 show that one step ahead CPU availability is relatively predictable. The slowly decaying autocorrelation between measurements means that ....

[Article contains additional citation context not shown here]

S. Gribble, S. Manku, D. Roselli, and E. Brewer. Self-similarity in file systems. available from http://www.cs.berkeley.edu/gribble.

....presence of a proxy cache, and if so, whether their behavior conforms to this intuitive explanation. One of the goals of this paper is to determine if this is indeed the case. Impact of burstiness on locality: Web workloads exhibit burstiness at multiple time scales (see Figures 1(b) and (c) [7, 11]. Periods of intense bursts in such workloads govern the tail of the load distribution. If the locality exhibited by the workload reduces during intense bursts, then the tail of the distribution will be relatively unaffected even in the presence of a proxy cache (since the reduced locality will ....

....we examine (a) whether proxies can help smooth out the burstiness in web workloads, and (b) the relationship between locality and the burstiness of web workloads. 3.3. 1 E ect of Caching on Burstiness Past studies have shown that web workloads exhibit burstiness at multiple time scales [7, 11]; the workloads employed in our study are no different (see Figures 1(b) and (c) In this section, we determine whether a proxy cache can help smooth out the burstiness in web workloads, or whether the resulting workload becomes more bursty. A better understanding of these issues will help ....

[Article contains additional citation context not shown here]

S. D. Gribble, G. Manku, D. Roselli, E. Brewer, T. Gibson, and E. Miller. Self-Similarity in File Systems. In Proceedings of ACM SIGMETRICS '98, Madison, WI, June 1998.

....tuning and benchmarking. Categories and subject descriptors: C.4 [Computer Systems Organization] performance of systems design studies, D.4.3 [Software] operating systems file systems management. 1 Introduction There is an extensive body of literature on usage patterns for file systems [1,5,9,11,14], and it has helped shape file system designs [8,13,17] that perform quite well. However, the world of computing has undergone major changes since the last usage study was performed in 1991; not only have computing and network capabilities increased beyond expectations, but the integration of ....

....impact on the network traffic; as for example the file size is a dominant factor in WWW session length. Given that the files and directories have heavy tailed size distributions, this directly results into heavy tailed distributions for those activities that depend on file system parameters [2,5]. Another important observation is that some characteristics of process activity, independent of the file system parameters, also play an important role in producing the heavy tailed access characteristics. From the analysis of our traces we find that process lifetime, the number of dynamic ....

[Article contains additional citation context not shown here]

Gribble, Steven, Gurmeet SinghManku, Drew Roselli, Eric A.Brewer, Timothy J.Gibson, and Ethan L.Miller; Self-similarity in file systems , in Proceedings of the SIGMETRICS'98 / PERFORMANCE'98 joint International Conference on Measurement and Modeling of Computer Systems, pages 141 -- 150, Madison, WI, June 1998.

.... function is suggestive of self similarity, and self similarity is often a manifestation of an unpredictable, chaotic series [5] Recent studies of network packet traffic [20] World Wide Web traffic [8] network protocol performance [22] transmitted video traffic [3] and networked file systems [17] all point to self similarity as an inherent property of modern distributed systems. Of particular interest is the work by Dinda and O Halloran [10] in which the authors rigorously analyze Unix load average data from a large number of computational settings. The focus of their analysis is on the ....

....The pervasiveness of these observations across our set of experiments supports the previous work of Dinda and O Halloran. We, therefore, surmise that the CPU availability exhibits long range autocorrelation and is either self similar (as noted in [10] or short term self similar as described in [17]. 3.2 Prediction of CPU Availability Over Longer Time frames Despite the long range autocorrelation present in CPU availability measurements, the data in Tables 2 and 3 show that one step ahead CPU availability is relatively predictable. The slowly decaying autocorrelation between measurements ....

[Article contains additional citation context not shown here]

S. Gribble, S. Manku, D. Roselli, and E. Brewer. Self-similarity in file systems. available from http://www.cs.berkeley.edu/~gribble.

No context found.

S. Gribble, G. Manku, D. Roselli, E. Brewer, T. Gibson, and E. Miller, "Self-Similarity in File Systems, Proceedings of ACM SIGMETRICS 1998.

....2.2.3 Handling Bursts and Incremental Growth As the load on the system grows, the SNS Manager must pull in idle nodes, and start new workers to deal with excess load. Moreover, the system must be able to deal with bursts in load. Network traffic has been shown to be bursty at varying time scales [37, 18, 33], and a network service must be able to handle such bursts. We deal with short traffic bursts by replicating workers and directing tasks across all replicated workers for greater throughput. More prolonged bursts, however, can result in stressing the system s resources. For example, after the ....

Gribble, S. D., Manku, G. S., Roselli, D., Brewer, E. A., Gibson, T. J., and Miller, E. L. Self-Similarity in File Systems. In Proceedings of ACM SIGMETRICS '98 (June 1998).

....local decisions regarding their choice of consumers for tasks. As the load on the system grows, the SNS Manager pulls in idle nodes, and starts new workers. In addition, the system must be able to deal with sudden bursts in load (Leland, Taqqu, Willinger Wilson 1994, Crovella Bestavros 1995, Gribble, Manku, Roselli, Brewer, Gibson Miller 1998). We deal with short traffic bursts by replicating workers and directing tasks across all replicated workers for greater throughput. To handle more prolonged bursts, our design supports the notion of a pool of overflow nodes that are not dedicated to the system, and are recruited by the system ....

Gribble, S. D., Manku, G. S., Roselli, D., Brewer, E. A., Gibson, T. J. & Miller, E. L. (1998), Self-Similarity in File Systems, in `Proceedings of ACM SIGMETRICS '98'.

No context found.

Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. ACM SIGMETRICS Conference on Measurement and Modeling of Computer Systems (Madison, WI, 21--26 June 1998.

No context found.

Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. In Proceedings of the 1998.

No context found.

Steven D. Gribble, Gurmeet Singh Manku, Drew Roselli, Eric A. Brewer, Timothy J. Gibson, and Ethan L. Miller. Self-similarity in file systems. ACM SIGMETRICS Conference on Measurement and Modeling of Computer Systems (Madison, WI, 21--26 June 1998.

No context found.

S. D. Gribble, G. S. Manku, D. Roselli, E. A. Brewer, T. J. Gibson, and E. L. Miller. Self-similarity in file systems. In Proc. ACM SIGMETRICS'98, pages 141--150, Madison, USA, June 1998.

No context found.

S. D. Gribble, G. S. Manku, D. Roselli, E. A. Brewer, T. J. Gibson, and E. L. Miller. Self-similarity in file systems. In Proc. ACM SIGMETRICS'98, pages 141--150, Madison, USA, June 1998.

No context found.

S. D. Gribble, G. S. Manku, D. Roselli, E. A. Brewer, T. J. Gibson, and E. L. Miller. Self-similarity in file systems. In Proceedings of SIGMETRICS, pages 141--150, 1998.

No context found.

S. D. Gribble, G. Manku, D. Roselli, E. Brewer, T. Gibson, and E. Miller. Self-Similarity in File Systems. In Proceedings of ACM SIGMETRICS '98, Madison, WI, June 1998. 25

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