Duncan H. Lawrie, J. M. Randal, and Richard R. Barton. "Experiments with automatic file migration." Computer, pages 45--55, July 1982.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....several companies had introduced automated tape systems [2] and automated tape storage became part of the mass storage systems in most large computer centers. Several of these centers were studied in the early 1980s; these included Brookhaven National Laboratory [5] the University of Illinois [10], and the Stanford Linear Accelerator Center [14,15] These will be discussed in a later section. Since these studies, many complex mass storage systems have been implemented, including those at NASA Ames, NCAR, and SDSC [7,11,12] However, no studies on these systems have been published. ....

....of 6.26 person minutes per day, given the file usage rates and the number of users on the system. For STP, this miss ratio would require a disk system that held 1.5 of the total tertiary storage, and would require 300 tracks, or about 1 MB, of data to be transferred each day. Lawrie, et al. in [10], considered the file migration patterns on the University of Illinois Cyber 175. Again, the system examined is quite different from the one studied in this paper. Interestingly, Lawrie reported that, though his system was quite different from SLAC, his results matched Smith s closely. This paper ....

Duncan H. Lawrie, J. M. Randal, and Richard R. Barton. "Experiments with automatic file migration." Computer, pages 45--55, July 1982.

....Previous Work Several studies on the reference patterns to mass storage systems have been published. Smith [Sm81d] analyzes file migration patterns in hierarchical storage management system. This analysis was used to design several HSM caching algorithms [Sm81c] Lawrie, Randal, and Burton [LRB82] compare the performance of several file caching algorithms. Miller and Katz have made two studies on the I O pattern of supercomputer applications. In [MK91] they find that much of the I O activity in a supercomputer system is due to checkpointing, and thus is very bursty. They make the ....

D.H. Lawrie, J.M. Randal, and R.R. Barton. Experiments with automatic file migration. Computer, pages 45-55, 1982.

....themselves are simply how a computer program chooses which files to move to free disk space. 3.1 Current Algorithms We examined four previously tested migration algorithms in this research. While we did not expect to discover anything beyond the information provided by Smith [29, 30] Lawrie [18], and Strange [32, 33] we wanted to corroborate our methodology and findings with their research. The algorithms we examined are first in, first out (FIFO) least recently used, LRU) size only, and space time. The FIFO algorithm is very simple and is a common algorithm better known for its ....

D. H. Lawrie, J. M. Randal, and R. R. Barton, "Experiments with Automatic File Migration," IEEE Computer, July 1982, pages 45--55.

....magazine and June, 1992 issue of SunExpert. effect of various file migration algorithms on the file system at the Stanford Linear Accelerator, and concluded that the best algorithm migrates files with the largest space time product, a quotient of their last reference time and file size. Lawrie [54] reached similar conclusions for a much different file system, the Illinois Cyper 175. More recently, Miller perfromed extensive evaluation of a modern massive storage system at the National Center for Atmospheric Research [71] He concluded that file migration policies should be designed to ....

Duncan H. Lawrie, J. M. Randal, and Richard R. Barton. Experiments with Automatic File Migration. Computer, pages 45--55, July 1982.

....our findings and outline directions for future work. 2 Related Work Though our understanding of input output parallelism is still evolving, there is a long history of file access characterization for mainframes and vector supercomputers. Notable examples include Lawrie and Randell s study [8] of automatic file migration algorithms, Stritter s analysis [25] of file lifetime distributions, Smith s study [24] of mainframe file access behavior, and Reed and Jensen s study [6] of file archive accesses. More recently, Miller and Katz [10] captured detailed traces of application file ....

Lawrie, D. H., Randal, J. M., and Barton, R. R. Experiments with Automatic File Migration. IEEE Computer (July 1982), 45--55.

....compressed segments to Footprint. RELATED WORK There has been some previous work concerned with automatic migration mechanisms and policies for tertiary storage management. Migration policy research has focused on whole file migration, with the primary exception of the Plan 9 work. Some studies [4,5] have determined that migration based solely on file size is inadequate. Strange [6] develops a migration model based on a daily clean up computation which migrates candidate files to tertiary storage once a day, based on the next day s projected need for consumable secondary storage space. ....

....future. So, for the time being, our migration decisions are on a per file basis, but in the near future they will be based on access ranges. There are a variety of migration benefit metrics: file size thresholds, time since lastaccess, and weighted combinations of the two metrics. Earlier studies [4,5] conclude that simply migrating the largest file, and preferentially retaining small files, does not work well for a scientific workload. Similarly, migrating the least recently accessed file may cause the system to migrate more files than necessary to reclaim space (e.g. if it migrates 100 1K ....

Lawrie, D.H., J.M. Randal, and R.R. Barton, "Experiments with automatic file migration," IEEE Computer, Vol 15, July 1982. pp. 45-55.

....that this policy would limit larger requests from evicting too many smaller requests from the cache. We call this policy SIZE MRU. The third policy we consider is a variation of SPACExAGE algorithm. SpacexAge algorithm is shown to be quite effective in minimizing misses in a mass storage system [8, 9]. In these systems, the product of size and age (since last use) of the file is used in determining the candidates for migration to tertiary storage. The variation we considered maintains a number of LRU chains instead of the usual single chain. The chains are partitioned based on request size. ....

D. H. Lawrie, J. M. Randal, and R. R. Barton. Experiments with automatic file migration. IEEE Computer magazine, pages 45--55, July 1982.

....of the system load. Several studies on the reference patterns to mass storage systems have been published. Smith [12] analyzes file migration patterns in hierarchical storage management system. This analysis was used to design several HSM caching algorithms [13] Lawrie, Randal, and Burton [7] compare the performance of several file caching algorithms. Miller and Katz 1 This work was performed while Theodore Johnson was an ASEE Summer Faculty Fellow at GSFC. This research is partially supported by grant from NASA through USRA, #5555 19 have made two studies on the I O pattern of ....

D.H. Lawrie, J.M. Randal, and R.R. Barton. Experiments with automatic file migration. Computer, pages 45--55, 1982.

....Previous Work Several studies on the reference patterns to mass storage systems have been published. Smith [Sm81d] analyzes file migration patterns in hierarchical storage management system. This analysis was used to design several HSM caching algorithms [Sm81c] Lawrie, Randal, and Burton [LRB82] compare the performance of several file caching algorithms. Miller and Katz have made two studies on the I O pattern of supercomputer applications. In [MK91] they find that much of the I O activity in a supercomputer system is due to checkpointing, and thus is very bursty. They make the ....

D.H. Lawrie, J.M. Randal, and R.R. Barton. Experiments with automatic file migration. Computer, pages 45-55, 1982.

....at GSFC. Previous Work Several studies on the reference patterns to mass storage systems have been published. Smith [14] analyzes file migration patterns in hierarchical storage management system. This analysis was used to design several HSM caching algorithms [15] Lawrie, Randal, and Burton [9] compare the performance of several file caching algorithms. Miller and Katz have made two studies on the I O pattern of supercomputer applications. In [11] they find that much of the I O activity in a supercomputer system is due to checkpointing, and thus is very bursty. They make the ....

D.H. Lawrie, J.M. Randal, and R.R. Barton. Experiments with automatic file migration. Computer, pages 45--55, 1982.

....the possible policy components discussed below require some additional mechanism support beyond that provided by the basic 4.4BSD LFS. They require some basic migration bookkeeping and data transfer mechanisms, which are described in the next section. 4.1. File Size based rankings Earlier studies [3, 12] conclude that file size alone does not work well for selecting files as migration candidates; they recommend using a space time product (STP) ranking metric (time since last access, raised to a small power, times file size) Strange [16] evaluated different variations on the STP scheme for a ....

D. H. Lawrie, J. M. Randal, and R. R. Barton. Experiments with Automatic File Migration. IEEE Computer, 15(7):45--55, July 1982.

....policy selects files for migration based on the interval since their last use, preferentially retaining active files on disk. This policy is based on locality of temporal reference: active files are likely to be accessed soon, while inactive files are not. Earlier studies by Lawrie et al. [3] and Smith [14] conclude that time since last access alone does not work well for selecting files as migration candidates; they recommend using a weighted space time product (STP) ranking metric, taking the time since last access, raised to a small power (possibly 1) times file size raised to a ....

D. H. Lawrie, J. M. Randal, and R. R. Barton. Experiments with Automatic File Migration. IEEE Computer, 15(7):45--55, July 1982.

....that this policy would limit larger requests from evicting too many smaller requests from the cache. We call this policy SIZE MRU. The third policy we consider is a variation of SpacexAge algorithm. SpacexAge algorithm is shown to be quite effective in minimizing misses in a mass storage system [8, 9]. In these systems, the product of size and age (since last use) of the file is used in determining the candidates for migration to tertiary storage. The variation we considered maintains a number of LRU chains instead of the usual single chain. The chains are partitioned based on request size. In ....

D. H. Lawrie, J. M. Randal, and R. R. Barton. Experiments with automatic file migration. IEEE Computer magazine, pages 45--55, July 1982.

No context found.

D.H. Lawrie, J.M. Randal, and R.R. Barton. Experiments with Automatic File Migration. IEEE Computer, 15(7), July 1982.

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