M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th Symposium on Operating Systems Principles, pages 35--50, Orcas Island, Washington, December 1985.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....objects. Even so, they do not fully exploit this potential, nor do they address the issue of protecting extensible systems: The access control mechanisms in Unix [16] are primitive and, barely, offer adequate security to protect file access. The access control mechanisms in the Andrew File System [19] and Windows NT [23] are more flexible than in Unix, but do not address the specific problems of protection in dynamically extensible systems as well. In this position paper, we identify the structure of extensible systems as it relates to access control and describe the current state of affairs. ....

....associates an individual and a group owner with each file [16] is primitive and barely sufficient for controlling file access, let al..one for controlling an extensible system. The Andrew File System uses fullblown access control lists, but does so only at the granularity of entire directories [19], which we believe is at too high a grain (see the discussion in section 2.3) Windows NT uses access control lists at the granularity of individual files and presents a rich, though unnecessarily complicated access control model [23] objects can be associated with three types of access ....

M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th Symposium on Operating Systems Principles, pages 35--50, Orcas Island, Washington, December 1985.

....to file blocks are provided from a set of file system instances, like in the Newcastle Connection [2] Systems which use the Sun NFS [31] protocol use caching of file blocks in local memory to improve performance. The Cedar file system [39] introduced caching of whole files. The Andrew file system [36] provides location transparency for files and adds read only file replication. The LOCUS [45] provides consistent sharing with a sophisticated locking and transactions mechanism for shared files. It also supports sharing of such files. This results in a complex interface and implementation. There ....

....in a stable state. For the scope of this report we are mostly concerned about client cache consistency. The other items are a property of the file server, the consistency guarantees in the client cache and the file server are orthogonal. Andrew Andrew is developed at Carnegie Mellon University [36]. The most important design assumptions is that the system is expected to scale to between 5.000 10.000 clients. There is a set of trusted servers called Vice that run the file server software. They form a shared name space, where each server has the responsibility for a subtree of the ....

Satyanarayanan, M., Howard, J. H., Nichols, D. A., Sidebotham, R. N., Spector, A. Z., and West, M. J. The ITC distributed file system: Principles and design. In Proceedings of the 10th Symposium on Operating System Principles (Orcas Island, WA, USA, December 1985), pp. 35--50.

....clients, the caches also reduce the communication delays that would otherwise be required to fetch blocks from servers. In addition, client caches reduce contention for the network and for the server machines. Since server CPUs appear to be the bottleneck in several existing network file systems [6, 14], client caching offers the possibility of greater system scalability as well as increased performance. There are two unusual aspects to the Sprite caching mechanism. The first is that Sprite guarantees workstations a consistent view of the data in the file system, even when multiple workstations ....

....from about 5 18 percent per active client to only about 1 9 percent per active client. Since normal users are rarely active, our measurements suggest that a single server should be able to support at least 50 clients. In comparison with Sun s Network File System [13] and the Andrew file system [14], Sprite completed a fileintensive benchmark 30 35 percent faster than the other systems. Sprite s server utilization was substantially less than NFS but more than Andrew. The rest of the paper is organized as follows: Section 2 gives a brief overview of Sprite; Section 3 describes prior work ....

[Article contains additional citation context not shown here]

SATYANARAYANAN, M., HOWARD, J. H., NICHOLS, D. A., SIDEBOTHAM, R. N., SPECTOR, A. Z., AND WEST, M.J. The ITC distributed file system: Principles and design. In Proceedings of the loth Symposium on Operating Systems Principles (Orcas Island, Wash., Dec. 1-4,

....sections. 2.2.1 Phase 1: Mobile File Systems Hosts in distributed computing environments often rely on the network in order to access shared resources, including the file system. Indeed, with distributed filing systems such as the Network File System (NFS) Sun,89] and Andrew File System (AFS) Satyanarayanan,85] it is possible for hosts to share files which are remotely located with other hosts. In the case of diskless workstations, all applications and data are implicitly located on remote filestore. Thus, when the network is unavailable or offering only a poor service quality, as is often the case in ....

....with access to files across a network. The overall architecture is based on the Andrew File System (AFS) which was developed at Carnegie Mellon University as part of the broader Andrew distributed computing environment and offers POSIX 1003.1 file system semantics [IEEE,94] Morris,86] Satyanarayanan,85] AFS supplies both the caching and management strategies while the basic file service is generally provided by Episode. The resultant file system provides a high performance, highly available, secure file repository and is tightly integrated with the DCE directory and security services. ....

[Article contains additional citation context not shown here]

M. Satyanarayanan, J. H. Howard, D. N. Nichols, R. N. Sidebotham, A. Z. Spector and M. J. West, "The ITC Distributed File System: Principles and Design", Proceedings of the loth Symposium on Operating System Principles (SOSP), Orcas Island, Washington, U.S., ACM Press, December 1985.

....the client is responsible for ordering bytes, managing network and server congestion, and otherwise handling the complex issues of implementing a distributed file system. This leaves the server simple and scalable [15] In fact, NFS servers maintain very little state. Satyanarayanan, et al. [16] justify this architecture by pointing out that in typical client server distributed systems, workstations have cycles to burn. Consequently, an NFS client tends to be complex, which can interfere with efficiency and correct behavior. Measuring NFS server performance is well understood. ....

Satyanarayanan, M., Howard, J., Nichols, D., Sidebotham, R., Spector, A., and West, M. "The ITC Distributed File System: Principles and Design." Proceedings of the 10th ACM Symposium on Operating System Principles, December 1985.

....to learn. Often this feeling that a construct would be much easier to create with the previous operating system bubble to the surface of consciousness. But there is also this new (sometimes scary) thought entering your mind: Wow, this might be useful, almost beautiful. 2 AFS and Coda AFS[8] was developed at Carnegie Mellon University (CMU) AFS was created as part of a prototype computing enviroment for Universities in a collaborative effort between IBM and Carnegie Mellon. Scalability is one of the main properties of AFS. It was designed to handle more then 5000 concurrent ....

Satyanarayanan, M., Howard, J.H., Nichols, D.A., Spector, A.Z., West, M.J. The ITC Distributed File System: Principles and Design, Proceedings of the Tenth ACM Symposium on Operating System Principles, Vol. 19, No. 5, Orcas Island, WA, USA, Dec. 1985.

....on the basis of NetBSD 1.2 and confirmed that the file system can run more e#ciently when the protection level is lowered. keywords: multi level protection, fail safe mechanism, distributed file system, operating system 1 INTRODUCTION A number of distributed file systems such as NFS [5] AFS [6], and Coda [7] have already been developed, and most of them are embedded in the monolithic kernel like UNIX systems. This makes it di#cult to develop distributed file systems because it is hard to debug the file systems in the kernel. The developers cannot use tools like a symbolic debugger that ....

M. Satyanarayanan, John H. Howard, D. A. Nichols, R. N. Sidebotham, A. Z. Spector, and M. J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th ACM Symposium on Operating Systems Principles, pages 35--50, 1985.

....subsystems. 4.5.1 File System Module The file systems are one of subsystems that are developed the most. Since the file systems a#ects the system performance largely, it is meaningful to improve the file systems and to develop new file systems. In particular, distributed file systems such as AFS [40] and Coda [41] are researched recently. CAPELA helps developers create such distributed file systems as well as local file systems. In the current implementation of CAPELA, the file systems are implemented on top of the virtual file system (VFS) 25] The file system modules serve the applications ....

Satyanarayanan, M., J. H. Howard, D. A. Nichols, R. N. Sidebotham, A. Z. Spector, and M. J. West, "The ITC Distributed File System: Principles and Design," in Proceedings of the 10th ACM Symposium on Operating Systems Principles, pp. 35--50, Dec. 1985. 57

....of storage. Systems designed to handle the storage and distribution of data within such archives must allocate, utilise and deallocate available resources in the most judicious manner, which can di#er drastically for di#ering sizes of data. For example, many network or distributed file systems [13, 31, 37] are aimed for use within typical operating environments by typical users. Studies [1, 34] and seemingly appropriate assumptions have been made regarding file size distributions, and designs tailored accordingly; however these assumptions may have no reasonable basis for large data archives. ....

M. Satyanarayanan et al. The ITC Distributed File System: Principles and Design. In Proc. 10th ACM Symp. on Distributed Systems Principles, volume 19, pages 35--50, December 1985.

.... storage systems [8] The challenges to the database community provided by trying to access tuples on tape was nicely summarized by Carey, Haas and Livny [3] The caching algorithm we propose is analogous to caching algorithms used for distributed file systems such as Sprite [26] and Andrew [28]. There have been a variety of caching algorithms for client server database system which utilize a page server architecture; see [21] and [10] for example. In these systems, a number of clients make requests for pages to a server. Our caching algorithm can be viewed as a multi level analogy to ....

....ptool is installed. Several different strategies are possible for the caching and migration of segments. Since segments are actually UNIX files, any distributed file caching system can be used, such as the one used in the Sprite file system [26] the 6 different 6. 2 Design Andrew file system [28], or the Apollo file system. The main design decisions are whether segments are cached to disk, as in Andrew, or to memory, as in Sprite; and whether segments are kept in a separate file cache, as in Sprite, or whether segments are managed using mapping and the virtual memory system, as in the ....

M. Satyanarayanan, J. H. Howard, D. A. Nichols, R. N. Sidebotham, A. Z. Spector, and M. J. West, "The ITC Distributed File System: Principles and Design," in ACM Press, New York, 1985, pp. 35--50. 16

....frequency. This paper proposes a distributed HLR approach for pointer forwarding, and compare its performance with the original (the single HLR) pointer forwarding scheme. 2 Pointer Forwarding with Single HLR Pointer forwarding is a well known technique used in Andrew distributed file system [14]. This technique was proposed to reduce the location update cost 4 VLR VLR VLR RA4 VLR VLR VLR RA1 RA2 RA3 RA6 RA5 PSTN HLR PSTN Figure 1: Pointer Forwarding with single HLR (the registration operation) in a PCS network [2, 4] In this algorithm, no message is sent to update HLR when a ....

Satyanarayanan, M., Howard, J.H., Nichols, D.N.,Sidebotham, R.N., Spector, A.Z., and West, M.J. The ITC Distributed File System: Principles and Design. Proc. 10th ACM Symposium on Operating Systems Principles, pages 35--50, December 1985.

....Keleher et al. 1992] Among these systems, we have chosen Coda for its use of logs in three different ways to resist site failure, network disconnection and partition. Coda [Kumar and Satyanarayanan 1991, Kistler and Satyanarayanan 1991] is a large scale distributed system, successor to Andrew [Satyanarayanan et al. 1985, Morris et al. 1986, Satyanarayanan 1987] From the logging viewpoint, Coda s originality stems from the use of operation logs, and the fact that these logs are placed in Recoverable Virtual Memory (RVM) a transactional memory itself logged in a value log. Coda uses logging in three different ....

Satyanarayanan, M., Howard, J. H., Nickols, D. A., Sidebotham, R., Spector, A. Z., and West, M. J. 1985. The ITC Distributed File System: Principles and Design. Proc. of the Tenth ACM Symp. on Oper. Syst. Principles, Orcas Island, WA (USA). Oper. Syst. Review, Vol. 19, N ffi 5, pp. 35--50. Dec.

....that they confront. The reader is strongly encouraged to read some of the original research papers on these and related systems. The Andrew File System was developed at Carnegie Mellon University and subsequently used as the basis of a world wide file system product offered by Transarc Inc [SHNS85, SS96]. The basic ideas in Andrew are easily summarized. AFS was built with the assumption that the Kerberos authentication technology would be available. We present Kerberos in Chapter 19, and hence limit ourselves to a brief summary of the basic features of the system here. At the time that a user ....

....of this work came a new generation of file systems that used closer cooperation between client and file system to exploit such patterns. Examples of well known file systems that take employ a stateful approach to provide increased performance (as opposed to availability) are AFS (see Figure 7 3) [Sat89, SHNS85, HKMN87] and Sprite[OCDN88, SM89] a research file system and operating system developed at U.C. Berkeley. On the availability side of the spectrum, the Coda project [KS91, MES95] a research effort at Carnegie Mellon University, takes these ideas one step further, integrating them into a file system ....

[Article contains additional citation context not shown here]

M. Satyanarayanan et al. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th ACM Symposium on Operating Systems Principles (Orcas Island, WA; Dec. 1985). ACM. 3550.

....such as the Newcastle Connection [5] provide direct access to the blocks of files from a named collection of file system instances. Performance is improved in the Apollo Domain file system [11] and Sun Microsystems NFS [22] by adding local caching of file blocks. The ITC distributed file system [15] adds location transparency for files and replication of read only files. It has adopted the transferring and caching of whole files used in CFS, but still maintains the traditional client model of shared, mutable files. The performance implications of this combination are not understood yet. In ....

Satyanarayanan, M., et al., "The ITC Distributed File System: Principles and Design," to appear in ACM SIGOPS Operating Sys. Review 19, 5 (Dec 1985). 21

....latest version of the file. 1.3.3. The ITC Distributed File System The Information Technology Center of Carnegie Mellon University is building a campus wide distributed system. Vice, the shared component of the distributed system, implements a distributed file system that allows sharing of files [28]. Each client workstation has a local disk, which is used for private files or shared files from a Vice file server. Shared files are copied as a whole to the local disk upon open, and the client operating system uses this local copy as a cache to satisfy disk requests. In this regard, the ITC ....

M. Satyanarayanan and John H. Howard and David A. Nichols and Robert N. Sidebotham and Alfred Z. Spector and Michael J. West. The ITC Distributed File System: Principles and Design. Operating Systems Review: Proceedings of the Tenth ACM Symposium on Operating Systems Principles 19(5):35-50, December, 1985.

.... Architecture of the RHODOS Distributed File System An analysis of the distributed file facilities described in [Braban and Schlenk 1989] Brown et al. 1985] Leach et al. 1985] Levine 1987] Mullender and Tanenbaum 1985] Fridrich and Older 1985] Sandberg 1986] Gifford et al. 1988] and [Satyanarayanan et al. 1985] show that their development is based on different architectural models. However, there are some common elements that allow to propose a general architecture, useful for research purposes. Such a general architecture has been proposed in [Goscinski 1991a] Here, we propose the architecture of a ....

....are very time consuming; it is necessary to avoid them whenever possible. Therefore to achieve the goal of high performance of the RHODOS system and motivated with the result of an analysis of the caching system presented in [Howard et al. 1988] Nelson et al. 1988] Ousterhout et al. 1988] [Satyanarayanan et al. 1985], and [Schroeder et al. 1985] we propose for RHODOS a caching system based on the client memory. The objective of the caching system for the RHODOS is to reduce the cost of file access by storing recently used blocks in local memory of file server or client s machine, and reuse them when it can be ....

Satyanarayanan, M., Howard, J.H., Nichols, D.A., Sidebotham, R.N., Spector, A.Z. and West, M.J. ITC Distributed File System: Principles and Design. Proceedings of the 10th Sympsium on Operating Systems Principles Orcas Island, pp. 35-50

....file system interface which allow mobility unaware applications to run without modification in a mobile environment. 3.4. 1 CODA The CODA file system [Satyanarayanan,90] is a highly available replicated file system developed from earlier work at Carnegie Mellon on the Andrew File System (AFS) Satyanarayanan,85] AFS is made up of stateful servers and one or more untrusted clients. The servers are physically secure, only run trusted software and are maintained by professional staff. The clients locally cache whole files from the servers; the consistency of cached files is maintained by the servers who ....

Satyanarayanan, M., J.H. Howard, D.N. Nichols, R.N. Sidebotham, A.Z. Spector, and M.J. West. "The ITC Distributed File System: Principles and Design." Proc. 10th Symposium on Operating System Principles (SOSP), Orcas Island, Washington, U.S., ACM Press, December 1985.

....of remote files. We currently simulate a global (i.e. laboratory wide) name space by following a set of conventions, and running update utilities we have written. This has worked reasonably well, but a distributed file system presenting a uniform name space to a set of machines, such as the ITC [6] and Sprite systems, is clearly a good idea and is something we plan to implement. We believe that the structure of such a global name space needs to permit individual users and groups of users to have control over their view of the name space. For example, in a software development environment, ....

M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC distributed file system: Principles and design. In Proceedings of the Tenth ACM Symposium on Operating System Principles, pages 35--50, New York, December 1985. ACM.

....campus wide and has approximately 350 Virtues on line with about 3600 registered users. Contact: Mahadev Satyanarayanan, Information Technology Center, Carnegie Mellon University, Schenley Park, Pittsburg, PA 15213 References: 33] 34] 35] 36] 37] 38] 39] 40] 41] 42] 43] [44], 45] 46] 47] 48] 49] 50] 51] 2.5 Argus Main Goal Argus is both a programming language for writing distributed applications as well as a system developed to support the implementation and execution of distributed environments. It copes with the problems that arise in distributed ....

M. Satyanarayanan, J.H. Howard, D.A. Nichols, R.N. Sidebotham, A.Z. Spector, and M.J. West, "The ITC Distributed File System: Principles and Design", In Proc. 10th Symp. on Operating Systems Principles, pages 35--50, ACM SIGOPS, December 1985.

No context found.

SAT 85 M. Satyanarayanan, J. H. Howard, D. A. Nichols, R. N. Sidebotham, and A. Z. Spector. The ITC Distributed File System: Principles and Design. Proceedings of the 10th ACM Symposium on Operating Systems Principles, 1985.

....does not meet the correctness criterion is at most . In Coda is composed of a probe interval of 10 minutes, and a message timeout of 15 seconds. A 0 correct protocol obeys the correctness criterion strictly. Even a validate on use protocol, such as in early versions of the Andrew File System [20] and Sprite [19] cannot achieve 0 correctness because of transmission delay. The timetable in Figure 3 shows the worst case behavior of a system whose failure detection interval is . We begin in the middle of a run, where C has f cached and a callback promise from S. Both C and S believe f C is ....

M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the Tenth ACM Symposium on Operating Systems Principles, pages 35--50, December 1-4 1985.

....Coda, Vnode Interface. 1. Introduction In this paper we describe and evaluate our approach for implementing a user level client cache manager on top of the Sun Vnode interface. Our work was carried out in the context of the Coda file system [5] a descendant of the Andrew file system (AFS) [4]. Coda is a highly available file system for a large scale distributed computing environment composed of Unix 1 workstations. It provides resiliency to server and network failures through the use of two distinct but complementary mechanisms. One mechanism, server replication, stores copies of a ....

Satyanarayanan, M., Howard, J., Nichols, D., Sidebotham, R., Spector, A., and West, M. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th ACM Symposium on Operating System Principles, Orcas Island (Dec. 1985).

No context found.

M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the Tenth ACM Symposium on Operating Systems Principles, pages 35--50, December 1-4 1985.

No context found.

M. Satyanarayanan, John H. Howard, David A. Nichols, Robert N. Sidebotham, Alfred Z. Spector, and Michael J. West. The ITC Distributed File System: Principles and Design. In Proceedings of the 10th Symposium on Operating Systems Principles, pages 35--50, Orcas Island, Washington, December 1985.

No context found.

M. Satyanarayanan, J. H. Howard, D. Nicholas, R. Sidebotham, A. Spector, and M. Vest. The ITC Distributed File System: Principles and Design. In Proc. 10th Symposium on Operating System Principles, pages 119130, Dec 1985.

First 50 documents  Next 50

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