K. K. Ramakrishnan, Prabuddha Biswas, and Ramakrishna Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of the 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....24. 8.3. Performance in More General Workloads Up to this point, the results in this section have applied only to workloads whose accesses are 100 small random writes. This section examines a mixed workload, defined in figure 26, modeled on statistics taken from an airline reservation system [Ramakrishnan92]. With this more general workload, the results of the earlier sections are modified by two important effects: reads and medium to large writes. The issues encountered in extending floating data and parity to handle variable sized access are complex and it is neglected from this section. For the ....

Ramakrishnan, K. K., et. al. Analysis of File I/O Traces in Commercial Computing Environments. In Performance Evaluation Review (SIGMETRICS). 20, 1 (1992). ACM, 78-90.

....file access patterns on 1 Unix workstations. Baker et al. 4] analyzed the user level file access patterns and caching behavior of the Sprite distributed file system. Smith [32] studied the file access patterns on IBM mainframes at the Stanford Linear Accelerator Center. Ramakrishnan et al. [28] performed a study of the I O access patterns in a commercial computing environment on a VAX VMS system. Jensen and Reed [17] studied the patterns of access to the file archive of the National Center for Supercomputing Applications. Miller and Katz [22] traced certain I O intensive applications on ....

K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of File I/O Traces in Commercial Computing Environments. In Proceedings of ACM SIGMETRICS and PERFORMANCE, pages 78--90, 1992.

....for sustaining period longer than 14.4 seconds is a ected by periodic behavior of hard drive. Is this value good enough As long as the re calibration period of HeadP redict module is longer than the average inter arrival time, most of the disk head prediction will be accurate. As pointed out in [15, 10], disk access usually falls into bursty pattern. In bursty period, the reference point is updated frequently. In sparse period, re calibration of HeadP redict updates the reference point timely before head prediction goes awry. So we conclude HeadP redict module works well in both cases. HeadP ....

K. K.Ramakrishnan, Prabuddha Biswas and Ramakrishna Karedla. Analysis of le I/O traces in commercial computing environments. In Proceedings of 1992 ACM SIGMETRICS and PERFORMANCE92, pages 78-89, June 1992.

....reported for Windows NT was for a development station, where in a short period a series of medium size files (5 8 Mb) containing precompiled header files, incremental linkage state and development support data, was read and written. 6. 2 File access patterns The BSD and Sprite (and also the VMS [15]) traces all concluded that most access to files is sequential. Summaries of our measurements, as found in table 3, support this conclusion for Windows NT file access, but there is also evidence of a shift towards more randomized access to files when compared to the Sprite results. A sequential ....

....tool for designing file systems and caches. There are 3 major tracing studies of general file systems: the BSD and Sprite studies [1,14] which were closely related and examined an academic environment. The 3 rd study examined in detail the file usage under VMS at a number of commercial sites [15]. One of our goals was to examine the Windows NT traces from an operating system perspective; as such we compared our results with those found in the BSD and Sprite studies. The VMS study focused more on the differences between the various usage types encountered, and a comparison with our traces, ....

Ramakrishnan, K. K., P. Biswas, and R. Karedla, Analysis of File I/O Traces in Commercial Computing Environments, in Proceedings of the 1992 ACM SIGMETRICS and Performance '92 International Conference on Measurement and Modeling of Computer Systems, pages 78-90, Pacific Grove, CA, June 1992.

....Several previous studies have analyzed the I O behavior from the system s point of view. Ganger and Patt presented a model for characterizing the system response for I O workloads [GP93] Application dependent file traces have also been extensively analyzed for studying the file system s response [RBK92, KN95] 3. Application characterization: The third approach instruments the I O intensive applications and characterizes their access patterns. Most previous work has focussed on characterizing I O behavior of scientific applications [CACR95] Previous studies of DSM machines have addressed ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of File I/O Traces in Commercial Computing Environments. In Proceedings of the SIGMETRICS'92/PERFORMANCE'92:Annual Conference on Measurement and Modeling of Computer Systems, pages 78--90, June 1992.

....by images. In the approximate range of 30,000 to 300,000 bytes, tail weight is increased mainly by audio files. Beyond 300,000 bytes, tail weight is increased mainly by video files. The fact that file size distributions have very long tails has been noted before, particularly in filesystem studies [1, 9, 17, 22, 24, 26], however they have not explicitly examined the tails for power law behavior and measurements of ff values have been absent. As an example, we compare the distribution of Web files found in our logs with an overall distribution of files found in a survey of Unix filesystems. While there is no ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of SIGMETRICS '92, pages 78--90, June 1992.

....of ff = 1:05; 1:95 traffic at 100s aggregation (left) 10s aggregation (middle) and 1s aggregation (right) Bottom row: Corresponding packet drop traces. a typical Unix file system indeed possesses a file size distribution that follows a power law and exhibits characteristics of heavy tailedness [35, 36, 25, 4, 33, 16]. This has also been confirmed in the case of reads and writes to NFS servers [6] and access patterns to documents on the World Wide Web [8] Caching, the use of thumbnails for images, and paging of large documents can certainly help dampen the degree of self similar burstiness by reducing the ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of the 1992 SIGMETRICS Conference on Measurement and Modeling of Computer Systems, pages 78--90, 1992.

....a file may be updated several times between pickups, but only a single version need be maintained in the cache. This further reduces the actual amount of client cache space required to store updates between pickups. File Lifetimes. Several studies have already been performed on file lifetimes [14, 39, 46, 49], and our own data confirms one of the earlier results: if a file is shortlived, it is usually very short lived. The majority of files that were both created and destroyed during our traces have a lifetime of less than five minutes. Furthermore, the great majority of these files live for less than ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of File I/O Traces in Commercial Computing Environments. ACM Performance Evaluation Review, 20(1):78--90, June 1992.

.... 85] measured isolated Unix machines. Baker et al. studied the workload of a cluster of workstations running the Sprite file system, which is a distributed, Unix like system [BHK 91] Ramakrishnan et al. studied access patterns in a commercial computing environment on a VAX VMS platform [RBK92] These studies all cover general purpose (engineering and office) workloads with uniprocessor applications. These studies identify several characteristics that are common among uniprocessor file system workloads: files tend to be small (only a few kilobytes) they tend to be accessed with small ....

K. K. Ramakrishnan, P. Biswas, and Ramakrishna Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of ACM SIGMETRICS and PERFORMANCE '92, pages 78--90, 1992.

....5 . 6.3. Performance in More General Workloads Up to this point, all of the analysis has been specialized for workloads whose accesses are 100 small (2KB) random writes. This section examines a mixed workload, defined in Figure 22, modeled on statistics taken from an airline reservation system [Ramakrishnan92] With this more general workload, the results of the earlier sections are modified by two important effects: reads and medium to large writes. The issues encountered in extending floating data and parity to handle variable sized access are beyond the scope of this paper and this technique is ....

....5 1 2 3 4 RAID level 5 4 3 2 1 With prereads Without prereads Nonredundant Parity Logging Floating Data Parity Mirroring 5 5 Fig. 22. Airline reservation workload. The I O distribution shown above was selected to agree with general statistics from an airline reservation system [Ramakrishnan92] This workload is reported as approximately 82 reads, a mean read size of 4.61 KB, and a median read size of 3 KB. The mean write size was larger, 5.71 KB, but the median write size was smaller, 1.5 KB. Locality of reference and overwrite percentages were not reported. All accesses are assumed ....

Ramakrishnan, K. K., Biswas, P., and Karedla, R. Analysis of file I/O traces in commercial computing environments. Proceedings of the 1992 ACM SIGMETRICS Conference on Performance. Newport RI, (June 1-5, 1992) published as a special issue of Performance Evaluation Review, ACM, 20, 1 (June 1992). ACM, 78-90.

No context found.

K. K. Ramakrishnan, Prabuddha Biswas, and Ramakrishna Karedla. Analysis of file I/O traces in commercial computing environments. Proceedings of 1992 ACM SIGMETRICS and PERFORMANCE92 International Conference on Measurement and Modeling of Computer Systems (Newport, RI). Published as Performance Evaluation Review 20(1):78--90, 1--5 June 1992.

No context found.

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of File I/O Traces in Commercial Computing Environments. In Proceedings of the 1992 ACM SIGMETRICS and Performance '92 International Conference on Measurement and Modeling of Computer Systems, pages 78--90, June 1992.

....Porcar [25] analyzed dynamic trace data for files in an IBM batch environment. Floyd and Ellis [12, 13] and Ousterhout et al. 23] studied file access patterns from isolated Unix workstations. Baker et al. 3] studied access pat terns in Sprite, a distributed Unix system. Ramakrishnan et al. [28] studied file access patterns in a commercial computing environment, on a VAX VMS platform. There have been a few studies of I O from scientific workloads. Del Rosario and Choudhary [10] provided an informal characterization of some grand challenge applications. Powell [26] concentrated mainly on ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of ACM SIGMETRICS and PERFORMANCE '92, pages 78--90, 1992.

....Porcar [25] analyzed dynamic trace data for files in an IBM batch environment. Floyd and Ellis [12, 13] and Ousterhout et al. 23] studied file access patterns from isolated Unix workstations. Baker et al. 3] studied access patterns in Sprite, a distributed Unix system. Ramakrishnan et al. [28] studied file access patterns in a commercial computing environment, on a VAX VMS platform. There have been a few studies of I O from scientific workloads. Del Rosario and Choudhary [10] provided an informal characterization of some grand challenge applications. Powell [26] concentrated mainly on ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of ACM SIGMETRICS and PERFORMANCE '92, pages 78--90, 1992.

....on performance. However, while contemporary file systems have been strongly influenced by the measurements reported in past workload studies, most past workload studies have been conducted in small scale academic distributed environments or in centralized commercial environments [1] 2] 3] 5] 6][9][13] This has led to a significant gap in our understanding of the behaviour of large scale distributed environments, and, in particular, large scale distributed industrial environments where usage patterns may differ significantly from those of academic environments. In this paper, we present ....

....reported by Floyd in an academic environment, as almost 60 of the read accesses were to files that had been read recently by 2 20 users on different workstations. Almost negligible read write sharing was reported. Studies of commercial environments have been analyzed by Ramakrishnan et al. [9]. They study eight different commercial VAX VMS environments in which the number of registered users (note that the Berkely study reported active users) in all the sites were in the range of 10 to 550. The sites covered different environments, including scientific research (200 users) program ....

[Article contains additional citation context not shown here]

K. K. Ramakrishnan, P. Biswas, R. Karedia, Analysis of File I/O Traces in Commercial Computing Environments, Proc. of ACM Sigmetrics and Performance'92 Conference, 20, 1, Newport, R.I., June 1992.

....has called engineering office applications [5] This workload applications using small files, with little concurrent sharing seems to be the standard workload assumed by most research file system designs. Other studies, even those that draw data from many different environments [7], confirm that this workload is very common. One notable exception is Echo [6] Through the use of read and write tokens, Echo provides replicated service and really does serialize reads and writes. Every reader sees the most recent data. This clean and strong guarantee comes at the cost of high ....

K. K. Ramakrishnan, P. Biswas and R. Karedla. Analysis of File I/O Traces in Commercial Computing Environments In Proc. 1992 ACM SIGMETRICS and PERFORMANCE '92, pages 78--90. ACM, June 1992.

....dynamic trace data for files in an IBM batch environment. Floyd and Ellis [Flo86, FE89] and Ousterhout et al. ODH 85] studied file access patterns from isolated Unix workstations. Baker et al. BHK 91] studied access patterns in Sprite, a distributed Unix system. Ramakrishnan et al. RBK92] studied file access patterns in a commercial computing environment, on a VAX VMS platform. The workload was dominated by office management and transaction processing applications. Studies of I O from scientific workloads have been fairly limited in scope and generality. Del Rosario and Choudhary ....

K. K. Ramakrishnan, P. Biswas, and Ramakrishna Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of ACM SIGMETRICS and PERFORMANCE '92, pages 78--90, 1992.

....as randomly sampling from such distributions. A definitive study of this question seems to not have been done. Previous measurement based studies of file systems have recognized that file size distributions possess long tails, but they have not explicitly examined the tails for power law behavior [24, 26, 19, 5, 23]. However, evidence of heavy tails in the distribution of file sizes has been noted in some specific contexts. In [9] it is shown that the size distribution of files found in the World Wide Web appears to be heavy tailed with ff approximately equal to 1. This result is also in general agreement ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of the 1992 SIGMETRICS Conference on Measurement and Modeling of Computer Systems, pages 78--90, June 1992.

....and whether file size access patterns can be modeled as randomly sampling from such distributions. Previous measurement based studies of file systems have recognized that file size distributions possess long tails, but they have not explicitly examined the tails for power law behavior [21, 22, 15, 4, 20]. In [7] it is shown that the size distribution of files found in the World Wide Web appears to be heavy tailed with ff approximately equal to 1, which stands in general agreement with measurements reported in [3] In [6] the authors show that the sizes of reads and writes to an NFS server ....

K. K. Ramakrishnan, P. Biswas, and R. Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of SIGMETRICS '92, pages 78--90, June 1992.

No context found.

K. K. Ramakrishnan, Prabuddha Biswas, and Ramakrishna Karedla. Analysis of file I/O traces in commercial computing environments. In Proceedings of the 1992.

No context found.

K. Ramakrishnan, P. Biswas, and R. Karedla, Analysis of file I/O traces in commercial computing environments, In Proc. of 1992.

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