O. Krieger and M. Stumm, "HFS: A flexible file system for largescale multiprocessors," in Proceedings of the DAGS/PC Symposium (The Second Annual Dartmouth Institute on Advanced Graduate Studies in Parallel Computation), 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of single machine architectures. Although HPF can reduce communication cost and, hence, increase the performance, this is only true for regular but not for irregular problems. Hurricane File System (HFS) The Hurricane File System is developed for large scale shared memory multicomputers [25]. HFS is a part of the Hurricane operating system. The file system consists of three user level system servers: the Name Server, Open File Server (OFS) and Block File Server (BFS) ffl The Name Sever manages the name space and is responsible for authenticating requests to open files. ffl The OFS ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for largescale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....that deal directly with I O at the level of files. The advantage of this approach is that the application programmer need not care for the basic problems of file I O since this is accomplished by the operating system, i.e. the file system. Examples of parallel file systems are Galley [10] HFS [11] , HiDIOS [19] ParFiSys[4] PIOUS [12] based on PVM [ and Vesta [8] Looking at one level above the file system, we come to the second category, namely the I O libraries. In short, this method enhances existing programming languages by features dealing with file I O. Thus, the application ....

Orran Krieger and Michael Stumm, HFS: a flexible file system for large-scale multiprocessors, Proceedings of the 1993 DAGS/PC Symposium (Hanover, NH), Dartmouth Institute for Advanced Graduate Studies, June 1993, pp. 6--14.

....database communities [10, 17, 25] We use mechanisms such as horizontal partitioning and two phase commits, but we do not need an SQL parser or a query optimization layer since we have no general purpose queries in our system. We also have much in common with distributed and parallel le systems [3, 23, 31, 33]. A DDS presents a higher level interface than a typical le system, and DDS operations are data structure speci c and atomically a ect entire elements. Our research has focused on scalability, availability, and consistency under high throughput, highly concurrent trac, which is a di erent focus ....

O. Krieger and M. Stumm. HFS: A Flexible File System for Large-Scale Multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6-14, Hanover, NH, Jun 1993.

....are thus an important factor in obtaining high throughput from both sequential and parallel file systems. Many studies have shown that file systems can utilize knowledge of input output access patterns to improve input output performance by selecting policies appropriate for the resource demands [26, 41, 68, 46]. Due to the complexity of input output optimization in parallel file systems, access pattern information is even more critical to performance than in sequential file systems. 1 In this thesis we propose an approach to optimizing parallel input output where automatic file access pattern ....

....Application Policy Control Given the natural variation in parallel input output patterns, tailoring file system policies to application requirements can provide better performance than a uniformly imposed set of strategies. Many studies have shown this under different workloads and environments [41, 46, 47, 26]. Small input output requests are best managed by aggregation, prefetching, caching, and write behind, though large requests are better served by streaming data directly to or from storage devices and application buffers. 11 There are several very closely related approaches to application policy ....

Krieger, O., and Stumm, M. HFS: A Flexible File System for large-scale Multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium (Hanover, NH, June 1993), Dartmouth Institute for Advanced Graduate Studies, pp. 6--14.

....file abstraction (a growable, addressable, linear sequence of bytes) with some parallel file access methods. The most common provide I O modes that specify whether and how parallel processes share a common file pointer [14, 21, 22, 23, 24, 25] Some systems are based on a memory mapped interface [26, 27], and two provide a way for the user to specify per process logical views of the file [28, 29] Some provide SIMDstyle transfers [30, 31, 25, 18] Finally, in addition to shared file pointers, MPI IO allows applications to describe a mapping from a linear file to the compute nodes running the ....

Orran Krieger and Michael Stumm, "HFS: a flexible file system for large-scale multiprocessors ", in Proceedings of the 1993 DAGS/PC Symposium, Hanover, NH, June 1993, Dartmouth Institute for Advanced Graduate Studies, pp. 6--14.

....File Systems Existing parallel I O models are often closely tied to the machine architecture as well as to the programming model. Typically jobs can access files in different I O modes , which determine how a file pointer is shared among clients running in individual nodes [7, 4, 18] The HFS [20] and KSR1 [17] file systems use a memory mapped interface. The nCUBE [9] and Vesta [5] allow more user control over data layout by providing per process logical views of the data. In PIFS [11] the file system controls which processor handles which part of the file to exploit memory locality. ....

O. Krieger and M. Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993.

....Existing parallel I O models are often closely tied to the machine architecture as well as to the programming model. Typically jobs can access files in different I O modes , which determine how a file pointer is shared among clients running on individual nodes [7, 4, 18, 14, 22] The Hurricane [20] and KSR1 [17] file systems use a memorymapped interface. The nCUBE [9] and Vesta [5] file systems allow more user control over data layout by providing per process logical views of the data. In PIFS (Bridge) 11] the file system controls which processor handles which part of the file to exploit ....

O. Krieger and M. Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993.

....configurable, malleable resource management algorithms provides a solution to this performance optimization conundrum. Below, we describe the two major components of this approach. 2.1. Classification Based Policy Selection Parallel file system research such as Patterson [4] Kotz [5] Krieger [6], and Grimshaw [7] has demonstrated the importance of tuning file system policies (e.g. caching, prefetching, writeback) to application access patterns. For example, access pattern information can be used to guide prefetching, small input output requests can be aggregated and large requests can ....

....of temporal variations. Given the natural variation in parallel input output patterns, tailoring file system policies to application requirements can provide better performance than a uniformly imposed set of strategies. Many studies have shown this under different workloads and environments [5,6,7]. Small input output requests are best managed by aggregation, prefetching, caching, and write behind, though large requests are better served by streaming data directly to or from storage devices and application buffers. There are several approaches to application policy control; these can be ....

KRIEGER, O., AND STUMM, M. HFS: A Flexible File System for Large-Scale Multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium (Hanover, NH, June 1993), Dartmouth Institute for Advanced Graduate Studies, pp. 6-14.

....input output patterns revealed by input output characterization studies [22] tailoring file system policies to application requirements can provide better performance than a uniformly imposed set of strategies. Many studies have demonstrated this under different workloads and environments [11, 13, 3]. For example, small input output requests are best managed by aggregation, prefetching, caching, and write behind, though large requests are better served by streaming data directly to or from storage devices and application buffers. Although applications with predictable resource demands can ....

Krieger, O., and Stumm, M. HFS: A Flexible File System for Large-scale Multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium (Hanover, NH, June 1993), Dartmouth Institute for Advanced Graduate Studies, pp. 6--14.

....fundamental idea of declustering file blocks across many disks for parallel access has been a common feature. Typically jobs can access files in different I O modes , which determine how a file pointer is shared among clients running in individual nodes [Cro89, BGST93, Kot93, Pie89, Roy93] HFS [KS93] and the KSR1 [Ken92] file system, use a memory mapped interface. On the nCUBE [DdR92] and in Vesta [CFPB93] the user has more control over data layout. These systems provide a per process logical view of the data. In PIFS (Parallel Interleaved File System) Dib90] the file system controls which ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....Most extend a traditional file abstraction (a growable, addressable sequence of bytes) with some parallel file access methods. The most common provide I O modes that specify whether and how parallel processes share a file pointer [7, 30, 33, 2, 17] Some are based on a memorymapped interface [23, 22]. Some provide a way for the user to specify per process logical views of the file [5, 9] Some provide SIMD style transfers [34, 24, 16] PIFS (Bridge) 11] allows the file system to control which processor handles which parts of the file, to encourage memory locality. Clearly, the industrial and ....

O. Krieger and M. Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....file abstraction (a growable, addressable sequence of bytes) with some parallel file access methods. The most common provide I O modes that specify whether and how parallel processes share a file pointer [Cro89, Pie89, Roy93, BGST93, Kot93] Some are based on a memory mapped interface [KSR92, KS93] Some provide a way for the user to specify per process logical views of the file [CFPB93, DdR92] Some provide SIMD style transfers [TMC87, Mas92, GGL93] PIFS (Bridge) Dib90] allows the file system to control which processor handles which parts of the file, to encourage memory locality. ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....files only via the external interface to the objects, applications are robust under architectural changes. 8.2. 6 HFS The Hurricane File System (HFS) is a part of the Hurricane operating system and was designed at the University of Toronto to run on the Hector shared memory multiprocessor [KS93, Kri94, SVW 93] While HFS shares our goal of providing a flexible, high performance file system, it adopts an entirely different approach. HFS is based on a complex, highly structured, object oriented model. Files in HFS are referred to as storage objects and are made up of three components, ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....in parallel input output patterns revealed by input output characterization studies [22] tailoring file system policies to application requirements can provide better performance than a uniformly imposed set of strategies. Many studies have shown this under different workloads and environments [11, 13, 3]. For example, small input output requests are best managed by aggregation, prefetching, caching, and write behind, though large requests are better served by streaming data directly to or from storage devices and application buffers. Although applications with predictable resource demands can ....

Krieger, O., and Stumm, M. HFS: A Flexible File System for Large-scale Multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium (Hanover, NH, June 1993), Dartmouth Institute for Advanced Graduate Studies, pp. 6--14.

.... new languages and compiler features that support I O parallelism and optimizations via data layout conversion [dBC93] compiler hints [PGS93] and preprocessing for out ofcore parallel code [CBH 94, CC94] Another approach integrates special enhancements for I O into the file system [CBF93, KS93] The theory community has developed parameterized computational models, called memory models or input output (I O) complexity models, that aim to represent the key features of computer memory hierarchies and data movement in order to present a suitable model for algorithm design and analysis at ....

Orran Krieger and Michael Stumm. HFS: A flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993.

No context found.

O. Krieger and M. Stumm, "HFS: A flexible file system for largescale multiprocessors," in Proceedings of the DAGS/PC Symposium (The Second Annual Dartmouth Institute on Advanced Graduate Studies in Parallel Computation), 1993.

No context found.

O. Krieger and M. Stumm, "HFS: A flexible file system for largescale multiprocessors," in Proceedings of the DAGS/PC Symposium (The Second Annual Dartmouth Institute on Advanced Graduate Studies in Parallel Computation), 1993.

....actively write shared are not replicated, but instead are accessed remotely. 4.4 The File System File I O in HURRICANE is provided largely outside of the kernel; only the device drivers are in the kernel. The HURRICANE File System (HFS) implementation makes use of three user level system servers [25, 26]. The Name Server manages the HURRICANE name space. The Open File Server (OFS) maintains the file system state for each open file, and is (largely) responsible for authenticating application requests to lower level servers. The Block File Server (BFS) controls the disks, determines the target for ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

....than those achieved by using mapped file I O for disk file accesses. The Alloc Stream Facility was originally developed as the application level library for the Hurricane file system, a file system being developed for shared memory multiprocessors with disks distributed across the multiprocessor [KS93] As part of this project, we are investigating different mechanisms for locking and managing the CIOS structures. Many of the file system policies, like pre fetching and compression are implemented by ASF in stream specific modules. Currently, ASF is being modified to: 1) allow the application ....

Orran Krieger and Michael Stumm. HFS: a flexible file system for large-scale multiprocessors. In Proceedings of the 1993 DAGS/PC Symposium, pages 6--14, Hanover, NH, June 1993. Dartmouth Institute for Advanced Graduate Studies.

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