Luis-Felipe Cabrera and Jim Wyllie. QuickSilver distributed file services: an architecture for hor- izontal growth. In Proceedings of the nd IEEE conference on computer workstations, Santa Clara, CA, March 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....The system is running in daily production use at IBM Almaden. Experiments are being made with other applications, including a messaging facility and a mail store and forward system. Contact: Roger Haskin, IBM Almaden Research Center, 650 Harry Road, San Jose, CA 95120 6099 References: [286], 287] 2.43 RFS Main goal Remote File Sharing (RFS) is one of the networking based features offered in the AT T UNIX system V.3, providing location transparent access to remote files and devices. Advantages RFS supports 100 of the UNIX file system s semantics. This means that, in ....

J.F. Cabrera and J. Wyllie, "QuickSilver Distributed File Services: An Architecture for Horizontal Growth", Technical Report RJ 5578 (56697), IBM Almaden Research Center, San Jose, January 1987.

....protocol the AM cooperates with the involved AMs on the other nodes as well as with the different participating DMs located on the same node. Fig. 8 shows the local part of the two phase commit protocol from the view of one participating node. 1 remote sites LogService AM AM site n A 2 6 4 5 Legend : 1 : initiate commit 2 : prepare request 3 : write prepare record for the data entries 4 : announce prepared 5 : write prepare ( commit) record A initiate distributed commit protocol 6 : write commit decision (commit done record) 7 : inform the DMs about the commit decision ....

....DMs located on the same node. Fig. 8 shows the local part of the two phase commit protocol from the view of one participating node. 1 remote sites LogService AM AM site n A 2 6 4 5 Legend : 1 : initiate commit 2 : prepare request 3 : write prepare record for the data entries 4 : announce prepared 5 : write prepare ( commit) record A initiate distributed commit protocol 6 : write commit decision (commit done record) 7 : inform the DMs about the commit decision 8 : write commit decision (committed done) for the data entries 9 : acknowledge commitment 10 : ....

[Article contains additional citation context not shown here]

L. F. Cabrera, J. Wylie : "Quicksilver Distributed File Services : An Architecture for Horizontal Growth", Proceedings of the 2th IEEE Conference on Computer Workstations, 23-26, Mrz 1989

....by operations invoked on the handle representing the item, independent of the naming framework. IV. SYSTEM OVERVIEW Several research projects have investigated naming facilities for distributed systems or systems with multiple servers, and have proposed a number of useful ideas and mechanisms[3, 4, 5, 6, 7, 8, 9, 10, 11]. This section presents an overview of the complete naming facility proposed for emulation systems under Mach 3.0, that tries to take advantage of the experiences from these earlier systems. A. Client Server Interactions and Access Mediation Servers give clients access to the items that they ....

L.-F. Cabrera and J. Wyllie, "QuickSilver distributed files services: An architecture for horizontal growth," Tech. Rep. RJ 5578, IBM Almaden Research Center, Apr. 1987.

....element to provide backward error recovery especially for distributed system architectures [7] 17] 32] The log is also used as a means to achieve persistence. There, the log is used for updating objects or files by atomically transferring their new created versions to permanent storage [4][9] Dependent on the application area, logs are used together with different representations of secondary, non volatile storage architectures. In transaction processing systems [1] 6] 23] usually data bases are used as permanent data repository. For many applications, where communication takes ....

L. F. Cabrera, J. Wylie : "Quicksilver Distributed File Services : An Architecture for Horizontal Growth", Proceedings of the 2th IEEE Conference on Computer Workstations, 23-26, März 1989

....an atomic file rename operation with the semantics that either the original copy of the file exists or it has been replaced by the new copy of the file. The solutions that have been used in other file systems to provide a higher measure of reliability than Sprite s are based on file versions [CaW86, SGN85] or atomic transactions [BKT85, PoW85] The systems that use file versions create a new version each time that a file is written. Thus, files will never be destroyed as a result of client or server crashes, because old versions of files will remain safely on disk. We chose not to use the version ....

L. F. Cabrera and J. Wylie, "QuickSilver Distributed File Services: An Architecture for Horizontal Growth", Research Report RJ 5578 (56697), San Jose, California, June 1986. 166

....for the file system prototype. Bill Schilit s labor produced the file traces used in the trace driven experiments described in this dissertation. vi 1 1 Introduction Distributed file systems are a fundamental structure of distributed computing, and much attention has been focused on their design [3, 9, 13, 18, 29, 30, 31, 44, 51, 53]. But one important new design point has not yet been explored thoroughly: how to support mobile clients. Portable workstations are becoming increasingly common, and people who use these machines should not be forced to accept inferior performance or usability. This leaves an open question for ....

....without the requirement that any two clients have precisely the same table. And because prefix tables contain only hints that need not be correct, this method avoids creating either an availability or a consistency problem. A very different approach to naming is taken in the QuickSilver system [9, 19], which employs the concept of user centered naming. Instead of all users having a single view of a global name space, each user has a logically distinct name space in which resolution is performed. It is possible to think of a global name space in which files are identified by (user, local ....

L.-F. Cabrera and J. Wyllie. QuickSilver Distributed File Services: An Architecture for Horizontal Growth. 117 In Proc. Second IEEE Conf. on Computer Workstations, pages 23--37, March 1988.

....It also contrasts our approach with related work. 4.1 Global vs. Per User Name Space Defining the file system on a per user basis seems to be well justified. The activity of accessing files by a single user tends to be isolated from other users, and focused on a small working set of directories [Cabr88, Floy86, Shel86]. Satyanarayanan [Saty89] has pointed out that in a research or academic environment, most files are read and written by a single user. When users share a file, it is usually the case that only one of them modifies it. This implies that file references outside the user s private name space are ....

....file system, but once the file system is mounted, the user can use the logical file system in a network transparent way. Finally, there are other distributed file systems that, like Jade, provide mechanisms to let users construct their own name spaces. Examples include Tilde [Come86] QuickSilver [Cabr88], and Plan 9 [Pres91] QuickSilver and Plan 9 do consider systems in large scale, but only in terms of size and wide area. Jade surpasses these systems in the ability to accommodate heterogeneity, allow for customization, and support interactions between name spaces. Generally, none of these three ....

Cabrera, L. F. and Wyllie, J. Quicksilver distributed file services: An architecture for horizontal growth. In Proceedings of the 2nd IEEE Conference on Computer Workstations, pages 23--37, Santa Clara, CA, March 1988.

....1 Introduction This work investigates how to maintain replicas in a distributed file system, especially one supporting mobile clients. While the topic of replica management within file systems has received so much attention that one might think there is no design point left unconsidered [3, 16, 14, 2, 9, 18, 1, 10, 4], the notion of mobile clients is a new development that alters operating circumstances and therefore suggests new designs. The idea of mobile file service clients stems from the exploding popularity of portable computers. We argue that existing file replica management schemes would not cope well ....

L. F. Cabrera and J. Wyllie. Quicksilver Distributed File Services: An Architecture for Horizontal Growth. Technical Report RJ5578, IBM Almaden Research Center, April 1987.

....given. Related Work File systems researchers have lately concentrated on providing new services to administrators and to users. Important areas of research include transaction protection, viewing past states of the file system ( time travel ) and attribute based naming strategies. QuickSilver [CABR88] is an early example of a file system that allows users to protect file changes with transactions. The Wisconsin Storage System (WiSS) CHOU85] was an early implementation of a storage manager supporting access to large data objects. WiSS decomposes large objects into pages, changes to which are ....

Cabrera, L., and Wyllie, J., "QuickSilver Distributed File Services: An Architecture for Horizontal Growth", Proc. 2nd IEEE Conference on Computer Workstations, March 1988.

....ad hoc recovery mechanisms are reimplemented within many applications. QuickSilver avoids this burden on applications by providing recovery in its file sys tem. The QuickSilver distributed file system (DFS) sup ports transactional access to files and directories on local and remote machines [3]. DFS guarantees that updates of committed transactions are safely on disk and will not be lost due to subsequent failures. When a transaction aborts, DFS undoes all changes made to the file system by the transaction. Thus, all files that were modified by the transaction are returned to their ....

Luis-Felipe Cabrera and Jim Wyllie. QuickSilver distributed file services: an architecture for hor- izontal growth. In Proceedings of the nd IEEE conference on computer workstations, Santa Clara, CA, March 1988.

....the copies of the storage mediator can improve performance of the system. The integrity of the storage mediator s data can be insured in several ways. One method is to let Swift administer the metadata specifying a high degree of resiliency. Another would be to use standard data base techniques [6]. 2.3 Storage Agents The storage agents administer all aspects of secondary storage media, including data layout optimization and off line data alignment. Each storage agent may administer many storage devices that can be disks, or other high speed devices including disk arrays. Since the Swift ....

L.-F. Cabrera and J. Wyllie, "QuickSilver distributed file services: An architecture for horizontal growth," in Proceedings of 2 nd IEEE Conference on Computer Workstations, (Santa Clara, California), IEEE Computer Society, March 1988.

....referenced by that directory. The integrity of the storage mediator s data can be insured in several ways. Our preferred method is to let Swift administer the metadata and specify a high degree of redundancy. Another approach would be to use standard data base techniques such as a write ahead log [14,15]. 2.3 Storage Agents The storage agents administer all aspects of secondary storage media, including data layout optimization and off line data alignment. Each storage agent may administer many storage devices. These storage devices can be disks or other high speed devices including arrays of ....

L.-F. Cabrera and J. Wyllie, "QuickSilver distributed file services: an architecture for horizontal growth," in Proceedings of the 2 nd IEEE conference on computer workstations, Santa Clara, CA, Mar. 1988.

No context found.

L. F. Cabrera,and J. Wyllie [1987]. QuickSilver Distributed File Services: An Architecture for Horizontal Growth. RJ5578, Computer Science Department, IBM Almaden Research Center, 1987.

No context found.

L.-F. Cabrera and J. Wyllie. QuickSilver distributed file services: An architecture for horizontal growth. In Proceedings of the 2nd IEEE Conference on Computer Workstations, pages 23--27, March 1988. 119

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