J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10:3-25, 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....some of the requirements that are particular to mobile use: any to any communication, larger replication factors, detailed controls over replication behavior, and the lack of pre motion actions. We omit discussion of well understood ideas, such as the case for optimistic replication, discussed in [2,3,5,17]. 2.1. Any to any communication By de nition, mobile users change their geographic location. As such, it cannot be predicted a priori what machines will be geographically collocated at any given time. Given that it is typically cheaper, faster, and more ecient to communicate with a local ....

....store a di erent subset of the ward set. Such replication exibility, called selective replication [10] provides improved eciency and resource utilization: ward members locally store only those objects that they actively require. Replication decisions can be made manually or with automated tools [5,8]. Since the ward set varies dynamically, di erent wards might store di erent sets: not all ward sets will be equivalent. In essence, the model provides selective replication between wards themselves. The reconciliation topologies and algorithms [10] apply equally well within a single ward and ....

J. J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1):3-25, 1992.

....subjected to higher error rates and larger bandwidth uctuations than xed networks. In addition, voluntary disconnections are often requested by users to reduce communication costs. Previous works in distributed systems have showed that caching of data is a central idea to disconnected operation [7, 8]. The reason is that cache can be used not only to improve performance but also to enhance availability. However, in mobile computing systems it is important to send to clients not only data but also code. In this way, it is possible to push to clients applications that can execute in the absence ....

J. J. Kistler and M. Satyanarayanan. Disconnected operation in the coda le system. ACM Transactions on Computer Systems, 10(1):3-25, Feb. 1992.

....Also, being wired for networking means that it is easy to provide power to a device, as it is already connected to cabling. However, the need to be physically attached to the network leads to a lack of mobility. One solution to this problem is the use of disconnection friendly le systems [13], which employ caching and other techniques to facilitate useful work whilst in a disconnected (and therefore mobile) state. However, this is not a complete solution, as such systems can only o er weak consistency. Wireless networking stands at the other end of the spectrum. Bandwidth is ....

James J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1):3-25, 1992.

....popular data items may warrant hundreds or thousands of replicas, while unpopular items may require no replicas. One dicult aspect of unconstrained replication is ensuring that content does not become stale. Slightly relaxed consistency, such as in the Web [10] OceanStore [16] or Coda [14], allows delay between the commitment of updates at the data source and the propagation of updates to replicas. None the less, update propagation must still occur in a timely manner. The potentially large number of replicas rules out direct, point to point delivery of updates to replicas. In fact, ....

J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1):3-25, February 1992.

....that manage additional time stamping data over an o the shelf le system or by the design of a le system extension that manages and hides the time stamp data. These tools can be used as a complement to the standard le system operations. 2 Related Work Replicated le systems such as Coda [4], Ficus [3] and Rumor [8] are bound to rely on some notion of replication volumes (typically subtrees in a given machine) These systems can be related to version control systems if interpreted as providing version control over a well de ned number of branches, when allowing optimistic replica ....

J. J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. In Thirteenth ACM Symposium on Operating Systems Principles, volume 25, pages 213-225, Asilomar Conference Center, Paci c Grove, US, 1991.

....pattern matching on the URL or hyperlink text, depth from the start document, and le properties such as size and mime type. This system basically requires a user to know and specify what documents he or she will need after disconnection; no attempt at speculation is made. The Coda lesystem[8] includes extensive provisions for caching les on a client for use when disconnected from the server; indeed one of the main 5 features of the Coda system is improved accessibility via caching, or hoarding as it is called in Coda. A Coda client caches les periodically or at user request, ....

James J. Kistler and M. Satyanarayanan. Disconnected operation in the coda le system. SOSP, 8(2):213-2251, October 1991. 18

....sparse. In practice, they can be distributed among many redistribution servers for scalability. Mechanisms to support this redistribution have been developed, but are not the focus of this paper. Using Hidden Markov Models has several bene ts over the systems evaluated in previous projects ([1, 2, 4, 5, 7]) Each of these projects attempted to measure pairwise relationships among les. This model, while simple and somewhat intuitive, does not take advantage of the intuition that le accesses are caused by some hidden task, which our two layer model attempts to detect. The rst advantage of HMM s, ....

J. J. Kistler and M. Satyanarayanan. Disconnected operation in the coda le system. ACM Transactions on Computer Systems, 10(1):3-25, Feb. 1992.

....of other updates, a given update has either completely nished or has not yet begun. Bayou supports a number of applications, including a group calendar, a shared bibliographic database, and a mail application [7] However, Bayou s consistency model applies more generally. For instance, Coda [15] provides speci c merge procedures for con icting updates of directories; this type of con ict resolution is easily supported under the Bayou model. With slight extensions, the model also supports Lotus Notes style con ict resolution, where unresolvable con icts result in a branch in the object s ....

....solution is to organize replicas into two tiers: a small number of base nodes which hold master copies of replicated objects, and a set of mobile replicas which may perform tentative updates. Gray s two tier scheme is similar in many ways to the organization of replicas in the Coda le system [15]. While Gray concentrated largely on performance issues, Coda explicitly addresses the di erences in resource availability, security, and integrity between rst class replicas hosted on servers and second class replicas hosted on desktop or mobile nodes. The second class replicas (cache copies) on ....

J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM TOCS, 10(1):3-25, February 1992.

....concurrency, other systems have tried to accommodate the resolution of con icts after they have occurred. These mechanisms are built into the system and allow for an application to de ne their resolution procedures based on application semantics. Examples of such systems are Coda, Bayou, and Rover [13, 6, 12]. Coda is speci cally designed for le systems. It uses an optimistic concurrency control scheme meaning that any client can update its copy of a piece of data. Coda automatically resolves concurrent updates to directories and also has a mechanism for transparent resolution of le con icts called ....

J. J. Kistler and M. Satyanarayanan. Disconnected operation in the coda le system. In Thirteenth ACM Symposium on Operating Systems Principles, pages 213-225, Asilomar Conference Center, Pacic Grove, U.S., 1991. ACM Press.

....more fragile relative to the wired medium [47] In a large distributed system, where communication costs are high relative to the cost of local storage accesses, cache consistency becomes the major problem. Prior work on le systems support mobility using weak consistency guarantees. Coda [54] and Ficus [81] for example, provide quick response time (availability) using optimistic replication [21] Disconnected AFS [44] and recent advances in Coda [73] exhibit the ecacy of partially connected operations to facilitate faster synchronization with servers on a backbone. However, these ....

....Limitations (ii) and (v) are related to the selection of data. The work in this report is mainly concerned with consistency of le systems in mobile computing environments. The feasibility of caching les has been well established for a totally disconnected operation by Ficus [39] and Coda [54]. Ficus and Coda assume that the user has access to intermittent high bandwidth to propagate the updates from the laptop computer to the backbone network. For achieving consistency, they use the techniques similar to those used for wireline systems like Andrew File System [43] The technique is to ....

[Article contains additional citation context not shown here]

J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1), February 1992.

....of laptop computers. Some of the challenges in using wireless networks for mobile computing are low bandwidth and high communication cost. These necessitate what is known as disconnected operation, i.e. using the laptop computer when network connectivity is not available. Researchers [KS92] have built systems in which the le system of a user on the backbone server 1 is replicated on the disk of a laptop computer. This kind of approach is known as optimistic replication. Studies [KS92, RHRP94] have found that usually there are very few con icting updates when using optimistic ....

....i.e. using the laptop computer when network connectivity is not available. Researchers [KS92] have built systems in which the le system of a user on the backbone server 1 is replicated on the disk of a laptop computer. This kind of approach is known as optimistic replication. Studies [KS92, RHRP94] have found that usually there are very few con icting updates when using optimistic replication for le systems. As such, optimistic replication is a good approach for disconnected operation. One of the problems of disconnected operation is that the replicated data on the backbone server ....

[Article contains additional citation context not shown here]

J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1), February 1992.

.... been successful for applications that operate on compound documents mainly because the complex and proprietary nature of these applications thwarts source code modi cations [9, 10] and the inclusion of several complex data types, usually embedded in a single le, makes system based adaptation hard [12, 18, 20]. For this paper, we studied compound documents generated by three popular applications of the Oce suite: Word, PowerPoint, and Excel. We chose to focus on Oce applications based on four factors. First, Oce is the most widely used productivity suite. Moreover, a signi cant number of Microsoft Oce ....

Kistler, J. J., and Satyanarayanan, M. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems 10, 1 (Feb. 1992), 3-25.

....Hoarding involves keeping a set of documents on the local machine so that, under conditions of minimal or nonexistent connectivity, this set of documents may be accessed and useful work can be accomplished. While the problem of hoarding has been approached both with respect to le systems [7, 9] and the Web [2] we feel that the latter domain has not been fully explored. In this project, we plan to study Web hoarding by writing a hoarding agent which incorporates user de ned priorities, recency of document use, and predicted access patterns to provide reasonable Web availability during ....

....most vital pages for o line browsing. All of these observations motivate the discovery of a system which will ensure that a set of important les is always close at hand, allowing useful work to be done regardless of the state of the network connection. The idea of hoarding is not a new one. In [7] the authors describe Coda, a le system designed to support disconnected access. While there are inherent di erences between le systems and the World Wide Web, the underlying hoarding mechanisms presented in Coda do apply to both environments. By combining implicit (recency of use) and explicit ....

James J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1):33-35, February 1992.

....They allow users to read or update the database while they are disconnected, merging the modi cations with other replicas when they reconnect [23, 58, 62, 63] Optimistic replication is the only choice in such environments. Applications with similar demands include mobile le systems (e.g. Coda [36, 52] and Roam [62] and mobile e mail systems (e.g. Bayou [71] and Lotus Notes [51] 1.3 Classifying optimistic replication algorithms We classify optimistic replication algorithms broadly along three axes: where an update can be issued, what is transferred as an update, and who transfers an ....

....update, retirement notices are circulated among the replicas to delete the update record. 2.1.3.2 Detecting update con icts Thomas write rule has one shortcoming: all con icting updates are silently discarded except for the newest one. Systems designed for multi user collaboration, such as Coda [36, 42], Roam [62] and Locus [57] need to know when updates con ict with one another to let users resolve the con icts. These applications combine timestamp vectors and contents transfer. They work almost like the systems based on Thomas write rule, except that they associate with each replica a ....

[Article contains additional citation context not shown here]

James J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Trans. on Computer Systems, 25(5), February 1992.

....[2] Distributed le systems such as xFS [3] provide strong ordering by ensuring that a single writer or multiple readers are able to access a le at a given time. Other systems permit weaker orderings to deal with disconnections or to improve performance or availability (e.g. Bayou [18] Coda [10]) Although orderings of writes is important to determine the value of a shared data item, in many interactive applications, how quickly write operations become e ective is also important. 2 For example, timeliness has been explored for dynamic content in the web [14] Ordering and timeliness ....

J. Kistler and M. Satyanarayanan. Disconnected operation in Coda le system. In ACM Symposium on Operating systems and Principles, 1992.

....autonomous and auto managed extension of the MU, dynamically customizable to the user needs and network status, that gives the agents more autonomy to work for long periods of time even when the MU is disconnected. The second line of related research exploits the use of proxies and software agents [6, 4, 5, 3, 7, 9]. To our knowledge, our work is the rst one to combine both aspects, by employing proxies and agents to provide users with storage external to their mobile computers. However, our work can bene t from other works, such as the CODA project [6] by implementing the ideas and techniques they propose ....

....the use of proxies and software agents [6, 4, 5, 3, 7, 9] To our knowledge, our work is the rst one to combine both aspects, by employing proxies and agents to provide users with storage external to their mobile computers. However, our work can bene t from other works, such as the CODA project [6], by implementing the ideas and techniques they propose for the maintenance of data consistency among copies stored in the MU cache, the GSN and the servers. The rest of this paper is structured as follows. In Sect. 2, we present the main features of the agents that participate in the service. In ....

J.J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10:3-25, 1992.

....little structure. Furthermore, it is inherently tied to a speci c processor architecture and does not tolerate failures, making it an unsuitable API for distribution outside of tightly managed installations. Files, which are clearly one of the most popular abstractions found in distributed systems [2, 8, 36, 37, 50, 56, 63, 67, 79, 85], represent a higher level of abstraction and o er a much larger granularity than memory. But, they also provide little structure, thus raising the issue of how to manage concurrent modi cation of the same le without relying on some form of external synchronization. As a result, les are an ....

....a simulation study show that the algorithms based on replication graphs considerably outperform the lock based protocol. The Oracle8i [58] database also uses two tier replication, but uses an optimistic strategy to asynchronously propagate transactions across rst tier nodes. The Coda le system [37, 68] uses two tier replication combined with an optimistic strategy to provide a scalable and highly available distributed le system. First tier nodes function as le servers and store replicated copies of les. In contrast to the database systems described above, which can execute transactions ....

James J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10(1):3-25, February 1992.

No context found.

J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10:3-25, 1992.

No context found.

J.J. Kistler and M. Satyanarayanan. Disconnected operation in the Coda le system. ACM Transactions on Computer Systems, 10:3-25, 1992.

No context found.

James J. Kistler and M. Satyanarayanan. Disconnected operation in the coda le system. ACM Transactions on Computer Systems, 10(1):3-25, February 1992.

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