S. Mishra and L. L. Peterson and R. D. Schlichting. A Membership Protocol based on Partial Order. Proceedings of International Working Conference on Dependable Computing for Critical Applications, February, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Section 6. 2 Related Work The membership problem was first defined for synchronous systems in [4] Since then it has also been the subject of intense investigation for asynchronous systems. There are two types of membership services for asynchronous systems: primary partition membership services [21, 17, 19, 18, 6, 15] and partitionable membership services [1, 16, 20, 7, 6] We use the timed asynchronous system model as the foundation of our work [5] This system model assumes that processes have access to a local unsynchronized hardware clock with a bounded drift rate. It uses the following failure model: ....

....model does not define any bound on the failure frequency. It is not only an accurate description of a network of workstations running Unix or Windows NT, it also describes many nearly synchronous systems accurately. Some of these membership services explicitly assume the time free system model [21, 19, 15], some assume the timed system model [16, 6] and for others it is not entirely clear if the underlying model is timed or time free. It was shown in [3] that the membership problem is not solvable in time free asynchronous systems. This impossibility result does not hold for timed systems [6] ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In J. F. Meyer and R. D. Schlichting, editors, Dependable Computing for Critical Applications 2, pages 309--331. Springer-Verlag, Wien, 1992.

....any process from another stable partition. Each stable process is required to have an up to date member set. 3 Related Work The membership service was first specified for synchronous systems in [3] It has also been specified and implemented for asynchronous systems by various authors, e.g. see [19, 18, 15]. In [2] the authors show that it is impossible to derive a useful specification for a membership service that is implementable in time free asynchronous systems without introducing extensions to the model like failure detectors. The reason for that impossibility result is that a subproblem common ....

MISHRA, S., PETERSON, L., AND SCHLICHTING, R. A membership protocol based on partial order. In Dependable Computing for Critical Applications 2, J. F. Meyer and R. D. Schlichting, Eds. Springer-Verlag, Wien, 1992, pp. 309--331.

.... to this trend can be found in [Reiter 1994] Examples of the close interrelationships between solutions to agreement problems and broadcast capabilities are the Advanced Automation System [Benel eta[ 1989, Cristian et 1990] Amoeba [Tanenbaum eta[ 1990, Kaashoek Tanenbaum 1991] Consul [Mishra et 1992, Mishra eta[ 1993] Delta 4 [Powell eta[ 1988, Powell 1994] Horus [van Renesse et al. 1995, van Renesse et al. 1996] Newtop [Ezhilchelvan et al. 1994] In fact, they represent systems where the service layers of broadcast communication and membership maintenance are intertwined together. 4. ....

S. Mishra, L. L. Peterson and R. D. Schlichting, "A Membership Protocol Based on Partial Order", in Dependable Computing for Critical Applications Two (J. F. Meyer and R. D. Schlichting, Eds.), pp.309-31, Springer-Verlag, Wien, Austria, 1992.

....is perceived to depart due to random communication delays. A member may depart from a group due to a normal shutdown, such as a scheduled maintenance, or due to a failure. The group membership problem has been studied extensively in the past both for synchronous and asynchronous systems, e.g. [9, 25, 29]. A detailed exposition of this problem is beyond the scope of this presentation. Informally, the strong group membership protocol, as described in [20] ensures that membership changes are seen in the same order by all members. In this protocol, a group of processors have a unique leader based ....

Shivakant Mishra, Larry L. Peterson, and Richard D. Schlichting. A membership protocol based on partial order. In Second Working Conference on Dependable Computing for Critical Applications, February 1990.

....operation basis, depending on the needs of the application and the costs they are prepared to pay for this guarantee. 13.10 The Group Membership Problem The role of a group membership service (GMS) is to maintain the membership of a distributed system on behalf of the processes that compose it [BJ87b, Cri91b, MPS91, MSMA91, RB91, ADKM92b, Gol92, Ric92, Ric93, RVR93, Mal94, MMA94, Aga94, BDGB94, Rei94b BG95, CS95, ACBM95, BDM95, FKMBD95, CHTC96, GS96]. As described above, processes join and leave the system dynamically over its lifetime. We will adopt a model in which processes wishing to join do so by first contacting the GMS, which updates the list of system members and then grants the request. Once admitted to the system, a process may ....

.... Three phase commit: Ske82b, Ske85] Byzantine agreement: Merxx, BE83, CASD85, COK86, CT90, Rab83, Sch84] Asynchronous Consensus: FLP85, CT91, CT92] but see also [BDM95, FKMBD95, GS96, Ric96] The method of Chandra and Toueg: CT91, CHT92, BDM95, Gue92, FKMB95, CHTC96] Group membership: [BJ87a, BJ87b, Cri91b, MPS91, MSMA91, RB91, CHTC96], see also [Gol92, Ric92, Ric93, RVR93, Aga94, BDGB94, Rei94b, BG95, CS95, ACBM95, BDM95, FKMBD95, CHTC96, GS96, Ric96] Partitionable membership [ADKM92b, MMA94] Failstop illusion: SM93] Token based total order: CM84, Kaa92] Lamport s method: Lam78b, BJ87b] Communication from non members ....

[Article contains additional citation context not shown here]

Shivakan Mishra, Larry L. Peterson and Richard D. Schlichting. A Membership Protocol Based on Partial Order. In Proceedings of the IEEE International Working Conference on Dependable Computing for Critical Applications (Feb 1991). 137145.

.... [MPS91b] Highways [Ahu93] Horus [vRHB94] Isis [BJ87] Maestro [BFHR98] Newtop [EMS95] Phoenix [MFSW95] Relacs [BDGB94] RMP [WMK95] Totem [AMMS 95] Transis [ADKM92b] and xAMp [RV92] group communication systems and the specifications in [SR93, MAMSA94, FvR95, RB91, CS95, Cri91, FLS97, MPS91a, HS95, JFR93, BDM95, BBD96, BDM97, DMS96, MS95, SS93, WS95, Nei96] 6.1 Assumptions Assumption 4.3.1 states that the network may not spontaneously generate messages and that all delivered messages are guaranteed to be uncorrupted. Note that this assumption does not restrict the network to ....

....(in the sense that several disjoint components of processes may co exist) or a primary partition membership service. Primary partition membership services are implemented in Isis [RB91] Phoenix [MS95] Consul [MPS91b] and xAMp [RV92] and discussed in the specifications of [Nei96] Cri91] and [MPS91a] While Consul [MPS91b] xAMp [RV92] and [Cri91] guarantee membership service properties only as long as no partitions occur, Isis [RB91] and Phoenix [MS95] do assume the possibility of partitions, but allow execution of the application to proceed only in a single partition. In Isis [RB91] ....

S. Mishra, L. L. Peterson, and R. D. Schlichting. A Membership Protocol based on Partial Order. In proc. of the intl. working conf. on Dependable Computing for Critical Applications, pages 137--145, Feb 1991.

....protocols that ensure processes agree on the current membership of the group. The group membership problem was first defined for synchronous systems by [Cri91] Since then, the group membership problem for asynchronous systems has also been the subject of intense investigation (e.g. KT91, MPS91, RB91, VVR92, JFR93, vRBC 93, BDGB94, BS94, DMS94, EMS95, MSMA94] Yet, despite the wide interest that it has attracted and the numerous publications on this subject, the group membership problem for asynchronous systems is far from being understood: In particular, there is no agreed ....

....a WGM protocol should not preclude the possibility that p is indeed the only process removed from the group. It is important to note that our result applies only to group membership services that attempt to maintain a single agreed view of the current membership of a group (e.g. RB91, KT91, MPS91, MSMA94, HS95] These are known as primary partition group membership services and are intended for systems with no network partitions, or for systems that allow the group membership to change in at most one network partition, the primary partition . So called partitionable group membership ....

Shivakant Mishra, Larry L. Peterson, and Richard D. Schlichting. A membership protocol based on partial order. In Proceedings of the IEEE International Working Conference on Dependable Computing For Critical Applications, pages 137--145, Tucson, AZ, February 1991.

....views. The group membership problem was first defined for synchronous systems by [Cri91] Since then, this problem has also been the subject of intense investigation for asynchronous systems. In particular, two types of group membership services have emerged: primarypartition, e.g. RB91, KT91, MPS91, MSMA94, HS95] and partitionable, e.g. ADKM92b, ADKM92a, JFR93, vRBC 93, BDGB94, EMS95] Roughly speaking, a primary partition group membership service maintains a single agreed view of the group (i.e. processes agree on their local views of the group) Such services are intended for ....

Shivakant Mishra, Larry L. Peterson, and Richard D. Schlichting. A membership protocol based on partial order. In Proceedings of the IEEE International Working Conference on Dependable Computing For Critical Applications, pages 137--145, Tucson, AZ, February 1991.

....vs4: If VM p delivers a view v, then for every component q in v, either VM q delivers v or VM p constructs consecutive view w that excludes q. view agreement. There are many protocols in the literature which can be used to implement the assumed VM subsystem; e.g. Birman87, Ricciardi91, Mishra91] for an asynchronous system with the primary partition assumption, Melliar Smith91, 5 Moser96, Amir92, Ezhilchelvan95, Babaoglu95] for partitionable asynchronous systems. These protocols are not designed to cope with g failures. The subsystem described below deals with g failures using the ....

....hand, is that a reconnected set need only contain a particular majority subset of last, never all last components as we would require in certain cases when a g failure occurs during view update (see example 3 of section 4. 2) The primary partition membership service in [Birman87, Ricciardi91, Mishra91] make the assumption that a majority of components in the Cview do not suspect each other and that a functioning component is rarely detected as failed. This assumption may not hold true during periods of network instability caused for example by bursty traffic or network congestion. This ....

S. Mishra, L. Peterson and R. Schlichting, "A membership Protocol Based on Partial Order", Proc. IFIP Conf. on Dependable Computing For Critical Applications, Tuscon, Feb. 1991, pp 137-145. 25

.... service is the primary component membership service of Isis [BvR94] It was followed by many other primary component membership services, for example, those of Phoenix [MS94] Consul [MPS91b] and xAMp [RV92] Primary component membership services are also specified in [CHTCB96, Nei96, Cri91, MPS91a, DPFLS98, DPFLS99] Consul [MPS91b] xAMp [RV92] and [Cri91] guarantee membership service properties only as long as no network partitions occur. In contrast, Isis [RB91] and Phoenix [MS94] do assume the possibility of network partitions, but allow execution of the application to proceed only in ....

S. Mishra, L. L. Peterson, and R. D. Schlichting. A Membership Protocol based on Partial Order. In Proc. of the intl. working conf. on Dependable Computing for Critical Applications, Feb 1991.

....This is known as following the primary partition model. Primary partition models propose to support applications in distributed asynchronous systems. CHTCB95] gives a formal specification on group membership and its impossibility in asynchronous systems [FLP85] Isis [Bir93] Amoeba [KT91] MPS91] or [RB91] are system examples of this model. They allow to manage strong consistency, but one of the disadvantages is that the system might block as long as the majority condition is not satisfied. Furthermore the system forces members of minority partitions to quit applications. Recently, many ....

....have received a message when we can remove it from the acknowledgment queue, but we do not guarantee that the same set of multicast messages is delivered in a given view. View ordering : another di#erence comes from the view change ordering. Most membership algorithms [Bir93] KT91] MPS91] RB91] follow the primary partition model. All require an unique ordered sequence of view changes of the membership. For the partitionable membership models [DMS94] BDMS97] all membership changes are totally ordered. Thus this membership algorithm guarantees virtual synchrony with total ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In Proceedings of the International Workshop on Parallele and distributed Algorithms, pages 137--145, February 1991.

....non arbitrary contextsj [59] Protocol : The speci cations of some means that allow a certain kind of interaction among a set of pieces of software. Otherwise stated, ithe set of rules that software components on dioeerent machines use to realize a given abstractionj [49]. Re nement : A process that renders more concrete an artifact. It is employed in the phrases iproperty re nementj, ispeci cation re nementj, and iarchitecture re nementj. The rst re nement applies on a property and produces a stronger one. The second applies on a pair of a ....

S. Mishra, L. L. Peterson, and R. D. Schlichting. A Membership Protocol Based on Partial Order. In J. F. Meyer and R. D. Schlichting, editors, Dependable Computing for Critical Applications 2, volume 6 of Dependable Computing and Fault-Tolerant Systems, pages 309331. Springer-Verlag Wien New York, 1992.

....replicas. Membership changes are also due to server failures (crashes) As it has already been mentioned (section 2) we assume the existence of a system failure detector [4, 18] which suspects and announces crashed replicas. The proposed architecture should accommodate a membership service [15, 18, 8], which will consistently interpret al..l the above configuration changes and detected failures into views [2] of membership, in the replicas. The decision about responsibility is a function on the current membership set, as it is perceived by a replica. Therefore, it is important for all ....

Shivakant Mishra, Larry L. Peterson, and Richard D. Schlichting. A membership protocol based on partial order. In Int. Working Conference on Dependable Computing for Critical Applications, Tuscon, AZ, February 1991.

....detections Figure 3: Components of the Membership Service 3.1. Failure Detector FD of member p (call it FDp) attempts to reach agreement over the suspicions reported by FSp, with the unsuspected members in the view. Agreement can be reached symmetrically involving all FDs [e.g. [5, 6]] or asymmetrically by electing one FD as the co ordinator [e.g. 7] Agreed suspicions become detections and the ids of detected members are passed on to VM for replacement. We describe two properties of FD (these can be met by many membership services published in the literature) We observe ....

.... current view and never suspect each other, will determine identical detected first set: o(active(p) active(q) p suspected q q suspected p ) view p (x) view q (x) d p (x) d q (x) d p (x) d q (x) The proof for FDP1 (in fact for a stronger version) is presented in [6]. Suppose that view p (x) is unique. Grouping members according to their detected first sets, will divide view p (x) into disjoint subsets of agreeing members. Let the subset that contains p be W p (x) W p (x) q view p (x) d p (x) d q (x) Each process r in view p (x) W p (x) is not in ....

[Article contains additional citation context not shown here]

S. Mishra, L. Peterson and R. Schlichting, "A membership Protocol Based on Partial Order", Proc. IFIP Conf. on Dependable Computing For Critical Applications, Tuscon, Feb. 1991, pp 137-145.

....It provides several multicast communication services among a group of processes. It provides the user with a history context of messages, and allows the user to define semanticaldependent multicast ordering. Failure notifications are ordered causally with respect to regular messages in the system [45]. The underlying message recovery protocol in Psync may be used on top of available hardware multicast. The Amoeba system contains support for ordered group multicast within the operating system kernel [34, 33] The ordered multicast protocol in Amoeba uses a centralized coordinator for ....

....that agreement cannot be reached forever even when no failures occur. On the other hand, the advantage of this approach is that membership changes are consistently ordered with respect to all regular messages. The Psync system possesses a membership mechanism based on causally ordered messages [45]. This membership protocol preserves causal order between membership changes and regular messages, but does not guarantee virtual synchrony. Jahanian et al. 31] provide a suite of membership protocols, unrelated to multicast message ordering. In the Weak Membership of [31] there is no guarantee ....

S. Mishra, L. L. Peterson, and R. D. Schlichting. A Membership Protocol based on Partial Order. In proc. of the intl. working conf. on Dependable Computing for Critical Applications, pages 137--145, Feb 1991.

....[19] are examples of such layers. UBLCS 97 08 2 Related Work For certain application classes with strong consistency requirements, it may be the case that all services have to be suspended completely in all but one partition. This situation corresponds to the so called primary partition model [31, 20, 26] that has traditionally characterized partitioned operation of network applications. In this paper we focus on system services that support partition awareness such that continued operation of network applications is not restricted to a single partition but may span multiple concurrent partitions. ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In Proceedings of the International Workshop on Parallel and Distributed Algorithms, pages 137--145, Tucson, AZ, February 1991.

No context found.

S. Mishra, L. L. Peterson and R. D. Schlichting, `A membership protocol based on partial order', in J. F. Meyer and R. D. Schlichting (eds), Dependable Computing for Critical Applications 2, SpringerVerlag, Vienna: 1992, pp. 309--331.

....is as hard as the group membership or the atomic broadcast problem [4] it also allows a deterministic solution of the group membership or the atomic broadcast problem. 4 Protocol Description 4. 1 Overview This protocol is based on the ideas developed in the membership protocols described in [20, 1, 21, 2]. Informally, the key idea in these protocols is that the team members exchange membership messages and each live team member p maintains a set S p of live team members based on these messages. The team members continue to exchange membership messages until the following property is satisfied: 8q ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In Proceedings of the Second Working Conference on Dependable Computing for Critical Applications, pages 137--145, Tucson, AZ, Feb 1991.

....protocol, an update is broadcast as soon as it is received from the application. This results in a much smaller delivery time in the Timewheel protocol, and also improves the stability time. The Timewheel membership protocol is different from other asynchronous membership protocols such as [21, 18, 1] in that it makes use of an internal clock synchronization service to detect process failures, elect a single decider in the system, and provide desired atomicity semantics. The use of clock synchronization service results in a much simpler design. In addition, the Timewheel membership protocol ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In J. F. Meyer and R. D. Schlichting, editors, Dependable Computing for Critical Applications 2, pages 309--331. Springer-Verlag, Wien, 1992.

.... divides the adaptation process into three different phases change detection, agreement, and action that can be used as a common means for describing various algorithms such as reliable transmission protocols and network services oriented towards faulttolerance such as group membership services [9, 19, 23, 27]. This serves not only to clarify the logical structure and relationship of such algorithms, but also to provide a unifying implementation framework. This paper is organized as follows. First, the general model of adaptive systems is described in Section 2. Section 3 then gives several examples ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In J. Meyer and R. Schlichting, editors, Dependable Computingfor Critical Applications2, pages 309--331.Springer-Verlag, Vienna, 1992.

....synchronous membership service [7] there is no bound on the time it takes to form a new group after failures or recoveries are detected. Protocols for implementing asynchronous membership services satisfying different properties for point to point and broadcast channel based networks are given in [14, 16, 18]. An asynchronous atomic broadcast service enables any member p of a group to broadcast an update u to all members of that group in such a way that u is delivered to all active and correct members in the same order. This service ensures the following property for any two team members p and q that ....

S. Mishra, L. Peterson, and R. Schlichting. A membership protocol based on partial order. In Proceedings of the Second Working Conference on Dependable Computing for Critical Applications, pages 137--145, Tucson, AZ, Feb 1991.

No context found.

S. Mishra and L. L. Peterson and R. D. Schlichting. A Membership Protocol based on Partial Order. Proceedings of International Working Conference on Dependable Computing for Critical Applications, February, 1991.

No context found.

M. Shivakant, L. Peterson, R. Schlichting. A Membership Protocol based on Partial Order. In Proc. of the IEEE Int. Working Conf. on Dep. Comp. for Critical Applications, pp 137-145, February, 1991.

No context found.

M. Shivakant, L. Peterson, R. Schlichting. A Membership Protocol based on Partial Order. In Proceedings of the IEEE Interbational Working Conference on Dependable Computing for Critical Applications, pp 137-145, February, 1991.

No context found.

MIPS91 Mishra S., Peterson L. L. and Schlichting R. D., A Membership Protocol Based on Partial Order, Proceedings of the IEEE International Working Conference on Dependable Computing for Critical Applications, Feb 1991, pp 137-145.

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