L. B. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365--379, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....disconnected op eration. 1.2 Existing Approaches This sub. 3ection describes current techniques for solving the hoarding problem in order to lay the groundwork for our own ideas. Do No hing. In their implementation of disconnected AFS [4] and their extensions for partially connected operation [5], Huston and Honeyman have not provided any special mechanisms for hoarding. Prior to disconnection, the user simply runs the applications she intends to use while traveling, thus filling the local disk cache appropriately. In addition to being inconvenient, this method suffers from an intrinsic ....

L. B. Huston and P. Honeyman. Partially Connected Operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing, pages 91-97, April 1995.

....on both serial links and LAN, using system configurations with and without our system. 1. 5 Our Contribution Previous work has focused on providing support for either semi connected operation (see [BB94] YB94] and [KRA94] or adaption to changes in the environment of a mobile host (see [HH95] and [NPS95] Our main contribution is providing the design and implementation of a system that integrates support for semi connected operation with support for adaption. Furthermore, we have enhanced the support for semi connected operation and adaption. The enhancements can be characterised as ....

....bandwidth allocated to the interactive traffic class. With a priority based general scheduler the function of the link sharing scheduler could be to reduce priorities of overbandwidth limit classes, causing these classes to be scheduled less frequently. 2.3. 3 Partially Connected Operation In [HH95] the problem of providing system performance in a MH operating semi connected 2 to those operating fully connected is addressed. As the bottleneck in a MH operating semi connected is the slow link used for interconnection, the focus here is on how to: ffl Limit traffic without hurting system ....

[Article contains additional citation context not shown here]

L. B. Huston and P. Honeyman. Partially Connected Operation. In Proceedings of the Second USENIX Symposium on Mobile and Location-Independent Computing, volume 8. The USENIX Association, Fall 1995.

....connection. Cache misses can be serviced, albeit slowly, and file updates can be sent back to the file server much more quickly, reducing the chance of update conflicts occurring. File systems with support for weakly connected operation have been described by Mummert (Coda) 8] and Huston (AFS) [6]. These systems are based on the same caching techniques which are used to support disconnected operation whole file caching of optimistically replicated files. Prior to going mobile the client s cache gathers files for use while weaklyconnected. This process is called hoarding in the Coda ....

....from a low technology connection, from network congestion, from distance from the home network, or from a combination of all three. 2 the log file as soon as possible in order to reduce the likelihood of update conflicts occurring due to sharing of files under an optimistic replication scheme [6, 8]. As well, performing write backs promptly helps reduce the size of the log file, which is desirable for freeing local client resources (disk space) as well as for reducing reintegration time once the client regains a strong connection [2] Write backs also help protect file modifications from ....

[Article contains additional citation context not shown here]

Huston, L., and Honeyman, P. Partially connected operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing (Ann Arbor, MI, April 1995), pp. 91--97.

....conflicts should any occur [11] Some studies have shown that update conflicts occur only rarely [2, 12, 7] and that optimistic replication is an acceptable strategy for a disconnected environment. 2.2. Weakly Connected Operation Weakly connected operation (also called partially connected [13]) is similar to disconnected operation, but with a few important extensions. A key advantage is that cache misses are no longer fatal. If the requested file is not in the cache, then the read request is sent to the home file server over the network, and the file is then transfered to the client ....

....back to the file server. Two conflicting goals complicate write back scheduling. First, it is desirable to write back the contents of the log file as soon as possible in order to reduce the likelihood of update conflicts occurring due to sharing of files under an optimistic replication scheme [13, 3]. As well, performing write backs promptly helps reduce the size of the log file, freeing local client resources (disk space) and the reintegration time. Timely write backs also help protect file modifications from loss due to failure, damage, or theft of the mobile client [3] Conflicting with ....

[Article contains additional citation context not shown here]

L. Huston and P. Honeyman. Partially connected operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing, pages 91--97, Ann Arbor, MI, April 1995.

....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 approaches still require intermittent high bandwidth connections or the synchronization cannot be completed quickly. Thus, the ....

....paper [77] some performance gures are provided for one to one reconciliation of email messages. Since the incoming email messages do not originate directly from a user s activity, process duplication is of little use for incoming email. 31 5.2. 5 University of Michigan work Huston and Honeyman [44] try to address the frustrations which a user is likely to experience as a result of the RPC latencies in a mobile environment. Their work is built on top of AFS [43] In AFS, a client opening a le is guaranteed to see the data stored when the latest (connected) writer closed the le. But in ....

[Article contains additional citation context not shown here]

L. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365-379, 1995.

....made on the laptop are not usually visible on the backbone server till the laptop is connected through a network to the server. Currently, a user of a laptop computer has to use her discretion in deciding when to connect to the server. This mode of operation is known as intermittent connectivity [HH95, MES95] Unfortunately, there do not exist any automated techniques to decide when to make the updates. Our model tries to ll this gap. The problem of accessing les over networks has been extensively studied by the theoretical computer science community. This has been mainly studied as a le ....

L. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365-379, 1995.

....needs to be built in at layers above the session layer and thus falls outside the scope of this work. Long term disconnections have been dealt with extensively in Coda and similar systems [35, 53] albeit in the limited context of file systems. We also do not deal with partial or weak connectivity [36, 61] which can essentially be treated as a combination of short term disconnections and reduced available bandwidth. We limit ourselves to dealing with mechanisms for handling short term and medium term disconnections at transport and RPC layers. Such mechanisms can be used by higher protocol layers, ....

....updates to such files and integrating the updates with the file server on reconnection. More recently, features have been added to Coda to handle weakly connected clients [61] exploiting links with low bandwidth and or intermittent connectivity, which are common in mobile computing. Little Work [35, 36] is another system that handles disconnected and weakly connected file system clients using file caching and logging of updates. The logged updates are propagated to the file server whenever possible in an environment with intermittent connectivity. Both the systems described above handle ....

L. Huston and P. Honeyman. Partially connected operation. In Proc. of the 2nd USENIX Symposium on Mobile and Location-Independent Computing, pages 11-- 24, April 1995.

....this file Trickle reintegration: an ongoing background process that propagates updates to servers asynchronously. Aging: a record is not eligible for reintegration until it spends a minimal amount of time, called aging window, in the log. Adaptive reintegration chunk size Little Work project [33] Background update propagation. Three levels of queueing priority in the network drivers: interactive traffic, other network traffic, and replay traffic. A number of tickets assigned to each queue. Immediate propagation of file updates to the client through callbacks. Directory updates are tricky, ....

L. B. Huston and P. Honeyman. Partially Connected Operation. Computing Systems, 4(8), Fall 1995.

....1 because communication is precluded when disconnected. Given that wireless wide area communication is possible from almost anywhere, the imposition of disconnection as the common untethered mode is artificial. While recent work has recognized the possible benefits of partial connection [6][8] 9] there exists no file system which treats partial connectivity as the common mode of operation. This paper describes the design and implementation of the PRAYER 1 file system (PFS) a mobility aware file system which optimizes for partial connectivity, and provides a generic interface ....

L. B. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365--379, 1995.

....leave the updating of the replicas to that server. Or, with other words, the user can select a optimistic or pessimistic approach. 3. Related work The challenges of systems that contain variably connected mobile nodes have been addressed in several projects like Coda [6] Ficus [3] LittleWork [5] and Bayou [10] Commercial systems and applications are available as well, for example Lotus Notes and Microsoft Access. They address the problems of disconnection and consistency from different points of view, and have found different balances between the conflicts regarding availability, mobile ....

L. B. Huston and P. Honeyman. Partially connected operation. In Proc. of the 2nd USENIX symposium on mobile and location-independent computing, pages 91--7. USENIX, 1995.

....file systems supporting disconnected operation, but relies on version vectors to detect conflicts [10] The Little Work project caches files to smooth disconnection from an AFS file system [11] Conflicts are detected and reported to the user. Little Work is also able to use low bandwidth networks [12]. The Bayou project [13] 14] defines a mobile aware database architecture for sharing data among mobile users. Bayou supports tentative operation logs and data values [15] and session guarantees for weaklyconsistent replicated data [16] To illustrate these concepts, the authors have built a ....

L. Huston and P. Honeyman, "Partially connected operation," in Proc. of the Second USENIX Symposium on Mobile & Location-Independent, Ann Arbor, MI, April 1995, pp. 91--97.

.... operation, that is the ability to operate disconnected, is essential for mobile clients [13, 14, 30, 25] In addition to disconnected operation, operation that exploits weak connectivity, that is connectivity provided by intermittent, low bandwidth, or expensive networks, is also desirable [20, 12]. Besides mobile computing, weak and intermittent connectivity also applies to computing using portable laptops. In this paradigm, clients operate disconnected most of the time, and occasionally connect through a wired telephone line or upon returning back to their working environment. Private or ....

L. B. Huston and P. Honeyman. Partially Connected Operation. Computing Systems, 4(8), Fall 1995.

....of disconnected operation. 1.2 Existing Approaches This subsection describes current techniques for solving the hoarding problem in order to lay the groundwork for our own ideas. Do Nothing. In their implementation of disconnected AFS [4] and their extensions for partially connected operation [5], Huston and Honeyman have not provided any special mechanisms for hoarding. Prior to disconnection, the user simply runs the applications she intends to use while traveling, thus filling the local disk cache appropriately. In addition to being inconvenient, this method suffers from an intrinsic ....

L. B. Huston and P. Honeyman. Partially Connected Operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing, pages 91--97, April 1995.

....can defer network communication until such a point that it is most likely to succeed. This policy is simple to implement using our event delivery abstractions, and can be very easily altered if necessary. Other environment dependent policies such as those described by Huston and Honeyman [12] can also be easily implemented using our event delivery abstractions. The event delivery abstractions proposed in this paper give an application the basic mechanism to be aware of the mobile computing environment. A major focus of our future work is to investigate and construct general mechanisms ....

L. B. Huston and P. Honeyman. Partially connected operation. In Proceedings of the 2nd USENIX Symposium on Mobile & LocationIndependent Computing, April 1995.

....useful research direction is the application and extension of the proposed proportional share mechanisms to manage resources not considered in this thesis. For example, lottery scheduling has recently been used to guarantee minimum service rates for scarce network bandwidth in mobile environments [HH95b] Power is another critical resource that can be affected by scheduling in mobile systems [WWDS94] It would be interesting to examine proportional share techniques for managing energy consumption. Another promising research area is the integrated use of proportional share scheduling techniques ....

Larry B. Huston and Peter Honeyman. Partially connected operation. In Proceedings of the Second USENIX Symposium on Mobile and Location-Independent Computing, pages 91--97, Ann Arbor, Michigan, April 1995.

....be be propagated back to the file server. Two conflicting goals complicate this problem. First, it is desirable to write back the contents of the log as soon as possible in order to reduce the likelihood of update conflicts occurring due to sharing of files under an optimistic replication scheme [6, 9]. As well, performing write backs promptly helps reduce the size of the log file, which is desirable for freeing local client resources (disk space) as well as for reducing reintegration time once the client regains a strong connection [3] Write backs also help protect file modifications from ....

....common for several modifications to be made to a file over a short period of time, and the number of write backs which need to be performed can be reduced by delaying the write back. Huston and Honeyman found that 70 of operations stored in a log could be eliminated through delaying write backs [6]. Ideally, we would like to gain the advantages of performing write backs without decreasing the performance of reads. Since these goals interfere with each other, the problem becomes one of trying to gain the most benefit from performing write backs with minimal interference to reads. Mummert, et ....

[Article contains additional citation context not shown here]

Huston, L., and Honeyman, P. Partially connected operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing (Ann Arbor, MI, April 1995), pp. 91--97.

....SQL procedures where you can define complex SQL commands and store them on the server [BP88] The stored commands are executed at the server end during a user transaction. Some distributed file systems support disconnected operation, including Coda [KS92, MES95] Ficus [RHR 94] and others [HH95] In these systems, applications on the laptop access the local file cache while the laptop is disconnected. On reconnection, the file system reconciles any differences with the appropriate file servers. The Bayou file system [TTP 95] internally uses a form of mobile code (but not agents) to ....

L. B. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365-- 379, Fall 1995.

....system also supports disconnected operation, but relies on version vectors to detect conflicts [39] The Little Work Project caches files to smooth disconnection from an AFS file system [19] Conflicts are detected and reported to the user. Little Work is also able to use low bandwidth networks [18]. These projects, as weel as Coda, are focused on hiding mobility from the application. The Bayou project [8, 45] defines an architecture for sharing data among mobile users. Bayou addresses the issues of tentative data values [47] and session guarantees for weakly consistent replicated data [44] ....

L. Huston and P. Honeyman. Partially connected operation. In Proc. of the Second USENIX Symposium on Mobile & Location-Independent Computing, pages 91--97, Ann Arbor, MI, April 1995.

....of operations is maintained. Upon reconnection, the operations in the log are played to the server after verifying that they would not result in conflicts. Support has also been developed for partially connected operation, a mode that can be used to improve performance over low bandwidth networks [7]. The Ficus distributed file system [15] developed at UCLA, also supports disconnected operation [3] Files may be replicated at any machine and accessed for reading and writing regardless of connectivity with other replicas. Periodically, replication sites communicate with each other to find out ....

L. Huston and P. Honeyman, "Partially Connected Operation," Second USENIX Symposium on Mobile and Location-Independent Computing, Apr. 1995, Ann Arbor, MI, pp. 91-97.

....file system supporting disconnected operation, but relies on version vectors to detect conflicts [16] The Little Work project caches files to smooth disconnection from an AFS file system [17] Conflicts are detected and reported to the user. Little Work is also able to use low bandwidth networks [18]. The BNU project implements an RPC driven mobile transparent application framework on mobile computers. It allows for function shipping by downloading Scheme functions for interpretation [19] The BNU environment includes mobile transparent proxies on stationary hosts for hiding the mobility of ....

L. Huston and P. Honeyman, "Partially connected operation," in Proc. of the Second USENIX Symposium on Mobile & Location-Independent Computing, Ann Arbor, MI, Apr. 1995, pp. 91--97.

....of the problem is that home and office have different ISPs. The two ISPs connect at the MAE West router on the west coast, so packets between New York City and northern New Jersey four miles as the crow flies travel across country and back again, typically going through 20 hops. e.g. [9, 14, 6, 8]) what I have implemented is unambitious and not technically groundbreaking. However, over time it became clear that operating at user level where tools such as file system utilities, scripting languages, etc. are available brings some profound advantages. My toolkit approach stands in ....

....Traveler s Briefcase. The other category is the published work on normal distributed file systems that have been extended to accommodate users who suffer various levels of disconnection. In this category are Coda [9, 10, 11, 13, 14, 15, 23] Ficus [3, 4, 18] and the Michigan alterations to AFS [6, 7, 8]. My work sits in between. The main difference with the commercial tools is that operation is more automatic, no special software is required on the master server, or both. Also, I know of no tools, commercial or otherwise, that use hierarchical checksums. The main difference with the research ....

[Article contains additional citation context not shown here]

L.B. Huston and P. Honeyman. Partially Connected Operation. In Proc. Second USENIX Symp. on Mobile and Location-Independent Computing Symp., USENIX, pages 91--97, April 1995.

....of existing mobile file systems. Finally, issues arising in file cache management in both stationary and mobile environments are discussed. 2.1 Motivation A significant amount of work has been done in developing new file system strategies to support mobile computing. Systems such as AFS [16, 17, 19], Ficus [14] and Coda [21] have all been adapted to provide file access for mobile clients. All of these systems rely on a file cache at the client for providing file availability. The performance of this cache has a major impact on the ability of a user to perform useful work. For example, if a ....

.... in its cache, the user may be stalled until the file can be retrieved which could take longer than a user is willing to wait (depending on the nature of the connection) While each file system has developed policies for managing certain aspects of the file cache (such as write backs in AFS [17, 19] and Coda [10] cache management policies have not been addressed in detail. This research examines the performance implications of various approaches to managing a mobile client s file cache. Two facets of cache management, the choices of hoarding and write back policies, are investigated in ....

[Article contains additional citation context not shown here]

L. Huston and P. Honeyman. Partially connected operation. In Proc. Second USENIX Symposium on Mobile and Location-Independent Computing, pages 91--97, Ann Arbor, MI, April 1995.

No context found.

L.B. Huston and P. Honeyman, "Partially Connected Operation," pp. 91-97 in Proc. 2nd USENIX Symp. on Mobile and Location-Independent Computing, Ann Arbor (April, 1995).

....The longer deferred updates sit in the cache, the more likely it is that this type of conflict will occur. On the other hand, by moving network dependencies out of the critical path, partially connected operation can significantly increase the speed at which mutating operations are performed [20]. Partially connected operation requires continuous network availability to support token based cache consistency protocols. This may be too expensive for users who pay for network connect time, or impossible in places that don t offer network access. Partially connected mode is especially ....

L.B. Huston and P. Honeyman, "Partially Connected Operation," pp. 91-97 in Proceedings of the Second USENIX Symposium on Mobile and Location-Independent Computing, Ann Arbor (April, 1995).

No context found.

L. B. Huston and P. Honeyman. Partially connected operation. Computing Systems, 8(4):365--379, 1995.

First 50 documents

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