I. Keidar. Moshe: A group membership service for wans, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....same subgroup and if this is not possible, ask the monitor to nd new locations to transfer the load. Our work is ongoing and we are currently implementing the two hierarchical layers over a real network environment. A recent attraction for researchers is implementing GCS over wide area networks [14]. We are also looking into using this architecture to implement a public key cryptography protocol over a wide area network. The protocol that is under consideration is the Die Hellman protocol [15] and the real time requirements are relaxed in this case. ....

I. Keidar et all, Moshe: A Group membership service for WANs, MIT Technical Memorendum MIT-LCS-TM-593a, 1999.

....into disconnected components of mutually connected members. Membership services strive to form and deliver the same views to all mutually connected members of the group. While this is not always possible, they typically succeed once network connectivity more or less stabilizes (see, for example, [58, 27]) Reliable multicast services allow application processes to send messages to the entire membership of a group. GCSs guarantee that message delivery satisfies certain properties. For example, one property can be that messages sent by the same sender are delivered in the order in which they were ....

....12, 8] continue executing their current synchronization protocol to termination and then deliver to the application a view that does not reflect the already detected changes in connectivity. Afterwards, the algorithm is invoked anew to incorporate the new changes. We refer to a view as obsolete [58] when it is delivered by a GCS even though the GCS already has information that the view s membership has changed. Obsolete views cause an overhead not just for the GCS, but also for applications. Since application processes do not know when the views delivered to them are obsolete, they handle ....

[Article contains additional citation context not shown here]

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for WANs. ACM Transactions on Computer Systems, 2002. To appear.

....the underlying system is that it is generally synchronous, but, rarely asynchronous. The GCS we propose follows from the French project ATR. 1 Reliable GCSs are powerful building blocks that facilitate the development of fault tolerant distributed systems and many such systems have been proposed [1, 10, 25, 8, 11, 14, 17, 21, 22, 24, 27]. But generally they are not designed for a real time context. A real time application has to ensure that tasks are made within some time limits. Accordingly, a real time GCS has to ensure a timeliness property: each message has a deadline and the messages have to be delivered by processes in ....

I. Keidar, J. Sussman, K. Marzull, and D. Dolev. Moshe: A group membership service for wans. Technical Report MIT-LCS-TM-593a, Laboratory for Computer Science, Massachusetts Institute of Technology, 2000.

....cost of a membership change. Maintaining an accurate group membership over a wide area network is dicult. Keidar et al. propose an algorithm to maintain group membership in a WAN but only demonstrate scaling to hundreds of clients and utilize a central server which CFS explicitly attempts to avoid [14]. Additionally, any large cache of node locations is likely to contain entries which have become unreachable (a linear group membership list could not be easily maintained) These unreachable entries will not a ect the correctness of a lookup but will cause the lookup to su er a large number of ....

Idit Keidar, Jeremy Sussman, Keith Marzullo, and Danny Dolev. Moshe: A group membership service for WANs. Technical Report MIT-LCS-TM-593a, 1999.

....(under, say, 50 nodes) under a single administrative domain, a RON can avoid the constraints imposed by scaling to Internet sizes, and the need to ignore attractive network paths. This observation leads to the basic design of a RON architecture. 3 Approach Our own experiments and other studies [15, 8] have shown that routing indirectly through other nodes often finds better communication paths than those provided by BGP. This observation drives us to develop a framework that allows small numbers of nodes to form an overlay that can take advantage of these improved paths. For deployment and ....

....write protocolspecific encapsulators, create databases to maintain and share performance information, or design protocols with which to transmit routing data. Peer to peer systems such as FreeNet [3] and Gnutella use ad hoc overlay networks for control or data transmission. The Moshe project [8] noticed that different statically configured overlays provided faster convergence than the default Internet routing. We believe that the development of many application level multicast systems can be eased with RONs. 4 When a cooperating RON system either controls the majority of the available ....

KEIDAR, I., SUSSMAN, J., MARZULLO, K., AND DOLEV, D. Moshe: A group membership service for wans. Tech. Rep. MIT-LCS-TM-593a, MIT, 2000.

....the same subgroup and if this is not possible, ask the monitor to nd new locations to transfer the load. Our work is ongoing and we are currently implementing two hierarchical layers over a real network environment. A recent attraction for researchers is implementing GCS over wide area networks [8]. We are also looking into using this architecture to implement a public key cryptography protocol over a wide area network. The protocol that is under consideration is the Die Hellman protocol [5] and the realtime requirements are relaxed in this case. ....

I. Keidar et all, Moshe: A Group membership service for WANs, MIT Technical Memorendum MIT-LCS-TM-593a, 1999.

No context found.

I. Keidar, J. B. Sussman, K. Marzullo, and D. Dolev. Moshe: A Group Membership Service for WANs. In ACM Transactions on Computer Systems, August 2002.

No context found.

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for WANs. ACM Transactions on Computer Systems, 20(3):1-48, August 2002.

No context found.

15 I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for WANs. ACM Trans. Comput. Syst., 20(3):1-48, August 2002.

.... has been shown useful for facilitating the design of distributed applications [16, 10] Our algorithm for implementing this semantics is more appropriate for WANs than the existing solutions: it requires less rounds of communication and is designed for the scalable WAN oriented architecture of [6, 31]. Our design is carried out at a very high level of formality and rigor, much higher than that of most previous designs of Virtually Synchronous GCSs. It includes formal and precise specifications, algorithms, and proofs. # This paper is an extended version of a paper, entitled A Client Server ....

....disconnected components of mutually connected members. Membership services strive to form and deliver the same views to all mutually connected members of the group. While this is not always possible, they typically succeed once network connectivity more or less stabilizes (see, for example, [31, 16]) In addition, GCSs provide reliable multicast services that allow application processes to send messages to the entire membership of a group. GCSs guarantee that message delivery satisfies certain properties. For example, one property can be that messages sent by the same sender are delivered ....

[Article contains additional citation context not shown here]

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev, Moshe: A group membership service for WANs, ACM Trans. Comput. Syst., (2002.

....over such metrics. One example is time complexity metrics. The typical theoretical metric used to analyze the running time of distributed algorithms is the number of communication rounds the algorithm performs, or the number of message exchange steps in case of a non synchronous system (e.g. [20, 14, 15]) In Section 3, we illustrate the weakness of this metric. Another example is reliability metrics. In [13] we highlight the fact that fault tolerant algorithms are often designed under the assumption that no more than t out of n processes or components can fail. This characterization of ....

....is to develop realistic ways to evaluate distributed algorithms. We hope that focusing on the right metrics will lead to the design of more e ective distributed algorithms and systems. Our research approach combines a range of research techniques: gathering of data [4] empirical evaluation [4, 15], and simulation [11, 17] as well as theoretical modeling and analysis [5, 6] We believe that these techniques complement each other, and when used together can lead to more e ective results. e.g. obtaining data on how real environments behave can lead to more realistic theoretical system ....

[Article contains additional citation context not shown here]

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for WANs. ACM Trans. Comput. Syst., 2002. To appear.

.... been shown to be useful for facilitating the design of distributed applications [16, 10] Our algorithm for implementing this semantics is more appropriate for WANs than the existing solutions: it requires fewer rounds of communication and is designed for the scalable WAN oriented architecture of [6, 31]. Our design is carried out at a very high level of formality and rigor, much higher than that of most previous designs of virtually synchronous GCSs. It includes formal and precise specifications, algorithms, # Received by the editors September 27, 2000; accepted for publication (in revised form) ....

....disconnected components of mutually connected members. Membership services strive to form and deliver the same views to all mutually connected members of the group. While this is not always possible, they typically succeed once network connectivity more or less stabilizes (see, for example, [31, 16]) In addition, GCSs provide reliable multicast services that allow application processes to send messages to the entire membership of a group. GCSs guarantee that message delivery satisfies certain properties. For example, one property can be that messages sent by the same sender are delivered ....

[Article contains additional citation context not shown here]

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev, Moshe: A group membership service for WANs, ACM Trans. Comput. Syst., 20 (2002), pp. 191--238.

No context found.

I. Keidar, J. B. Sussman, K. Marzullo, and D. Dolev. Moshe: A Group Membership Service for WANs. In ACM Transactions on Computer Systems, August 2002.

No context found.

I. Keidar. Moshe: A group membership service for wans, 1999.

No context found.

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for wans. ACM Transactions on Computer Systems, 20(3):1-48, 2002.

No context found.

I. Keidar, J. Sussman, K. Marzullo, and D. Dolev. Moshe: A group membership service for wans. ACM Transactions on Computer Systems, 20(3):1-48, 2002.

No context found.

I. Keidar et at. Moshe: A group membership service for wans, 1999.

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