Marshall K. McKusick, Michael J. Karels and Keith Bostic, A Pageable Memory Based Filesystem, In Proc. Summer 1990 USENIX Conference, pg. 137-143, Anaheim, CA, June 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Case 4: Leases, Attrib leases, Rdirlookup The NQNFS protocol using leases for cache consistency,with the readdir RPC replaced by the Readdir and Lookup, and requiring a valid lease not only for file data access, but also for file attribute access. 2 Testing using the 4. 4BSD MFS [McKusick90] resulted in slightly degraded performance, probably since the machines only had 16Mbytes of memory,and so paging increased. As can be seen in figure 1, the buffer cache achieves about optimal performance for the range of two to ten megabytes in size. At eleven megabytes in size, the ....

Marshall K. McKusick, Michael J. Karels and Keith Bostic, A Pageable Memory Based Filesystem, In Proc. Summer 1990 USENIX Conference, pg. 137-143, Anaheim, CA, June 1990.

....amortize the cost of seeking between segments anywhere on the disk, while flash cleaning is a requirement of the hardware. Finally, Kawaguchi et al. 11] recently designed a flash memory file system for UNIX based on LFS, with performance comparable to the 4. 4BSD Pageable Memory based File System [17]. They found that at low utilizations cleaning overhead did not significantly affect performance, but the throughput of random writes dropped from 222 Kbytes s to 40 Kbytes s when utilization increased from 30 to 90 . A reduction of 82 in write throughput is comparable to 28 Chapter 1 the ....

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A pageable memory based file system. In USENIX Conference Proceedings, pages 137--144, Anaheim, CA, Summer 1990. USENIX.

....the improvement is more significant for MAC and DOS than for HP. Only the HP trace significantly benefits from an SRAM cache larger than 32 Kbytes. Disks were spun down after 5s of inactivity. The size of the DRAM buffer cache was 2 Mbytes for MAC and DOS and and no DRAM was used for HP. tem [16]. They found that cleaning overhead did not significantly affect performance, but they need more experience with cleaning under heavier loads. 7 Conclusions In this paper we have examined three alternatives for file storage on mobile computers: a magnetic disk, a flash disk emulator, and a flash ....

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A pageable memory based file system. In USENIX Conference Proceedings, pages 137--144, Anaheim, CA, Summer 1990. USENIX.

....the processor s write buffers along with the PCI SCI adapter s streams. 3 Related Work SciOS draws on the research in multiple areas. The use of memory based file systems have traditionally been used to increase performance of access to temporary files. Both the Pageable Memory Based File system [11] and tmpfs [12] implement memory based file systems. SciOS further extends this concept by providing a distributed memory based file system. The Global Memory Service (GMS) 5] and the NChance forwarding system [3] provide mechanism to use idle remote memory as swap space. In SciOS, we combine ....

Marshall K. McKusick, Michael K. Karels, and Keith Bostic. A pageable memory based filesystem. In Proceedings of the Summer 1990 USENIX Technical Conference, pages 137--143, June 1990.

....et al. show in [48] by comparing the 4.4BSD log structured file system (which is based on ideas first explored in Sprite LFS [42, 44] and the BSD fast file system [29, 41, 27] that the highest performing data organization is highly workload dependent. Temporary, memory based file systems [30, 35] have been added to many systems because conventional file creation deletion semantics are too heavy weight for some applications, costing more than a factor of two in application performance [16] All of these examples show that storage management should be tailored to the needs of individual ....

....do not need to use the same persistence model for all files, but can use different persistence models for different files. The concept of allowing different files to have different integrity and persistence characteristics is not new. For example, temporary and memory based file systems [35, 30] were added to systems so that files that are known to be short lived do not have to suffer the high cost of file creation and deletion imposed by conventional file systems. One major difference between these approaches and the exodisk system model is that they partition the files with specific ....

[Article contains additional citation context not shown here]

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A pageable memory based filesystem. In Proceedings of 1990 Summer USENIX, pages 137--143, Annaheim, California, June 1990.

....to amortize the cost of seeking between segments anywhere on the disk, while flash cleaning is a requirement of the hardware. Kawaguchi et al. 10] recently designed a flash memory file system for UNIX based on LFS, with performance comparable to the 4. 4BSD Pageable Memory based File System [16]. They found that cleaning overhead did not significantly affect performance, but they need more experience with cleaning under heavier loads. 7 Conclusions In this paper we have examined three alternatives for file storage on mobile computers: a magnetic disk, a flash disk emulator, and a flash ....

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A pageable memory based file system. In USENIX Conference Proceedings, pages 137--144, Anaheim, CA, Summer 1990. USENIX.

....unlike special purpose devices. The goal of the Rio (RAM I O) file cache is to achieve the performance of main memory with the reliability of disk: write back performance with write through reliability. We achieve memory performance by eliminating all reliability induced writes to disk [McKusick90, Ohta90]. We achieve reliability by protecting memory during a crash and restoring it during a reboot (a warm reboot) Extensive crash tests show that even without protection, warm reboot enables memory to achieve reliability close to that of a writethrough file system while performing 20 times faster. ....

....in [Ganger94] Rio still performs 8 faster than this system due to the sync every 30 seconds. Yet while these systems lose 30 seconds of recently written data on a crash, Rio loses none. MFS, which is completely memory resident and does no disk I O, is shown to illustrate optimal performance [McKusick90]. Though suitable only for temporary files, MFS achieves superior performance because of its simplicity its code is 1 10 the size of UFS s Adding protection slows Rio performance down by 20 , but some applications may require the extra margin of safety this provides. 5 Architectural Support ....

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A Pageable Memory Based Filesystem. In Proceedings USENIX Summer Conference, June 1990.

....for the buffer cache, the memory device driver, or memory mapped files. 6 Related Work Nearly all modern file systems use a file buffer cache to speed up disk accesses [Nelson88, Leffler89] Many have a memory file system that stores a complete, though temporary, file system in memory [McKusick90 ]. To our knowledge, however, the only file system that attempts to make its files reliable while in memory is Phoenix [Gait90] Phoenix keeps two versions of an in memory file system. One of these versions is kept write protected; the other version is unprotected and evolves from the ....

Marshall Kirk McKusick, Michael J. Karels, and Keith Bostic. A Pageable Memory Based Filesystem. In Proceedings USENIX Summer Conference, June 1990.

....the FFS code and put in a new, generic file system module (UFS) This code contains all the directory traversal operations, almost all vnode operations, the inode hash table manipulation, quotas, and locking. The common code is used not only by the FFS and BSD LFS, but by the memory file system [MCKU90] as well. The FFS and BSDLFS implementations remain responsible for disk allocation and actual I O. In moving code from the FFS implementation into the generic UFS area, it was necessary to add seven new vnode and vfs operations. Table 4 lists the operations that were added to facilitate this ....

Marshall Kirk McKusick, Michael J. Karels, Keith Bostic, "A Pageable Memory Based Filesystem," Proceedings of the 1990 Summer Usenix Technical Conference, Anaheim, CA, June 1990, 137-144.

....This code contains all the directory traversal operations, almost all vnode operations, the inode hash 8 1993 Winter USENIX January 25 29, 1993 San Diego, CA table manipulation, quotas, and locking. The common code is used not only by the FFS and BSDLFS, but by the memory file system [MCKU90] as well. The FFS and BSD LFS implementations remain responsible for disk allocation, layout, and actual I O. In moving code from the FFS implementation into the generic UFS area, it was necessary to add seven new vnode and vfs operations. Table 4 lists the operations that were added to facilitate ....

Marshall Kirk McKusick, Michael J. Karels, Keith Bostic, "A Pageable Memory Based Filesystem," Proceedings of the 1990 Summer Usenix Technical Conference, Anaheim, CA, June 1990, 137-144.

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