Lamport L: On Interprocess Communications: I. Basic Formalism. Distributed Computing, 1: 77-85 (1986)  Home/Search   Document Details and Download   Summary   Related Articles   Check

....we present some examples of hierarchical communication systems for various architectures and address the problem of supplying the agreed multicast operation in these systems. 47 7. 2 A Model Concurrency in hierarchical systems has been the subject of considerable attention in the literature ([20, 19, 22]) especially in the context of data base systems ( 25, 3] In this section we present a model for hierarchical systems using the concurrent object paradigm introduced in [20] and [19] A concurrent object is a simplified form of I O automata. A full description of the concurrent object paradigm ....

....processes and objects. A concurrent system is composed of an arbitrary set of concurrent objects. A concurrent object is both a client, invoking a sequential stream of requests to other objects, and a server, serving other objects requests. ffl We do not use an auxiliary global time order (see [22]) but rather use a precedence function. A precedence function maps histories to precedence relations and is part of the system s specification. We now turn to give a precise definition to the notion of an hierarchical concurrent system. 7.2.1 Hierarchical Concurrent Systems A system object O is ....

L. Lamport. On interprocess communication. Distributed Computing, 1(2):77--85, 86-- 101, 1986.

.... C rest 1 (2.11) 8i# t :0 6= i N ; 1 x i 6=0 p x i t p i : CS i (t) IH (t) c i c j t) 2.12) 38 2. 4 Lock Free Object Implementations Lock free shared objects have been proposed as viable alternatives to lock based objects in general asynchronous systems byvarious researchers [1, 3, 4, 36, 37, 49, 54, 67, 68, 74]. In this dissertation, weusetheterm lock free to refer to object implementations based on an unbounded retry loop structure like that depicted in Figure 1.2. Some lockfree implementations do not adhere to this characterization. For example, there exists an important special class of lock free ....

....refer to object implementations based on an unbounded retry loop structure like that depicted in Figure 1.2. Some lockfree implementations do not adhere to this characterization. For example, there exists an important special class of lock free implementations known as wait free implementations [36, 37, 67, 54] in which operations must satisfy a strong form of lock freedom that precludes all waiting dependencies among tasks, including potentially unbounded retry loops. Formally, a shared object implementation is lock free iff the following holds: if several objects access a shared object concurrently ....

[Article contains additional citation context not shown here]

L. Lamport. On interprocess communication, parts 1 and 2. Distributed Computing, 1:77--101, 1986.

....client issuing a write may not know when the write completes. This is not a problem from a theoretical standpoint, since this knowledge is required by neither safe semantics (provided by the SBQ protocol for generic data) nor regular semantics (provided by the SBQ protocol for self verifying data) [16]. Furthermore, completion of write operations is both well defined from the point of view of an observer external to the system, and timely, in the sense that completion cannot be delayed by faulty servers because it only depends on actions taken by correct processes. Nonetheless, SBQ protocols do ....

L. Lamport. On interprocess communications. Distributed Computing, pages 77--101, 1986.

....cooperation among concurrent processes We believe the answer to be no. To take an example from nature, it is doubtful that bees making up a swarm have individual names, yet, they cooperate e#ectively in performing highly complex tasks. The key to communication is not naming but, as Lamport [20] points out, the existence of a persistent communication medium (the beehive, the intruding bear, the bees themselves) and a coherent interpretation of the information it encodes. In the programming language arena, there are numerous instances where data access is primarily by content rather than ....

L. Lamport. On interprocess communication. Distributed Computing, 1:97--111, 1986.

....it is based on the observation that in every execution of every optimal algorithm one can identify a tree, consisting of edges over which certain 22 important messages were sent. These messages, termed causal, are messages which have a causal relationship (viz. Lamports happened before relation [L86]) to the output of the computed function at node 1. An algorithm may send many causal messages, but if we consider for every node the edge over which it sent the last causal message, this defines a tree. This tree can then be used by a tree based algorithm. See the appendix for more details. We ....

L. Lamport, On Interprocess Communication, Pat I, II. Distributed Computing, Vol 1, pp. 77-101, 1986.

....which initializes the component and precedes all other operations on it. A run on a composite register construction is atomic or linearizable, if the partial order on its operations can be extended to a strict total order #, such that for any scan s and for each component k it holds that [13]: 1. # k #s# # s and 2. there is no update u on X k such that # k #s# # u # s. A construction is atomic if all its runs are atomic. When sub registers are atomic, the precedence relation is a to Cwfscan 1094#s C receive scan 256#s C lfscan 72#s C send msg 10#s C send scan 240#s C receive ....

L. LAMPORT (1986) On interprocess communication, part i: basic formalism, part ii: basic algorithms. Distributed Computing 1, 77101.

....also write key for name and key range for name space. Interprocess Communication. We use interprocess communication through shared memory and allow arbitrarily initialized shared memory (dirty memory model) as in [19] Shared memory primitives such as wait free atomic read write registers [17, 18] are widely used in the theory of distributed algorithms [12] A deterministic protocol executed by n processes is wait free if there is a finite function f such that every non faulty process terminates its protocol executing a number f (n) of steps regardless of the other processes execution ....

....computation is usually ignored. We use waitfree atomic read write registers as primitives. Such primitives must be ultimately implemented wait free from singlereader single writer wait free atomic read write bits (that in turn are implementable from mathematical versions of hardware flip flops [17]) The most efficient such implementations use [18] to reduce a multiuser multivalue register to single reader single writer multivalue registers, and [17] to reduce the latter to single reader single writer bits. To standardize complexity and to make comparisons to other algorithms unambiguous we ....

[Article contains additional citation context not shown here]

Lamport L: On Interprocess Communication. Distrib Comput 1:86--101 (1986)

....[17, 20, 14, 3] address non atomic communication operations in the context of self stabilization. Lamport initially demonstrated that interprocess communication without explicity synchronization is possible [15] and formalizations of less than atomic communication were subsequently developed in [21, 16]. The register hierarchy and register constructions of [16] inspired and active research area. The register hierarchy (safe, regular, and atomic registers) has many motivations, including implementation cost for shared register operations. Another view of weaker forms of registers (safe or ....

....the context of self stabilization. Lamport initially demonstrated that interprocess communication without explicity synchronization is possible [15] and formalizations of less than atomic communication were subsequently developed in [21, 16] The register hierarchy and register constructions of [16] inspired and active research area. The register hierarchy (safe, regular, and atomic registers) has many motivations, including implementation cost for shared register operations. Another view of weaker forms of registers (safe or regular, when compared to atomic) is that they are possible ....

[Article contains additional citation context not shown here]

L. Lamport, \On interprocess communication, parts 1 and 2," Distributed Computing, vol. 1(1), 1986, pp. 77-101.

....need was one of the primary motivations for this work. We have developed a set of definitions and theorems that allow us to reason about the class of shared variables implemented by quorum systems. We have adapted Lamport s formal definitions of the concepts of safe, regular, and atomic semantics [2], which have traditionally been used to describe the semantics of Byzantine quorum systems, so that they can be applied in a direct and formal manner to such variables. Using these formalisms, we have proved that the writeback mechanism used in [4] to produce atomic semantics for dissemination ....

....ordered by timestamp, and this total order is consistent with the partial order in which the operations are actually processed. Following the example of [3] we discuss the semantics of these variables using a slightly generalized version of Lamport s definitions of regular and atomic semantics ([2]) We omit our formal specification of the generalized properties for reasons of space. Informally, however: Definition 1. A TS variable v is regular if any read operation on v returns either the value of the most recently completed write operation or the value of some write operation that is ....

L. Lamport. On interprocess communications. Distributed Computing, 1:77--101, 1986.

....client issuing a write may not know when the write completes. This is not a problem from a theoretical standpoint, since this knowledge is required by neither safe semantics (provided by the SBQ protocol for generic data) nor regular semantics (provided by the SBQ protocol for self verifying data) [16]. Furthermore, completion of write operations is both well defined from the point of view of an observer external to the system, and timely, in the sense that completion cannot be delayed by faulty servers because it only depends on actions taken by correct processes. Nonetheless, SBQ protocols do ....

L. Lamport. On interprocess communications. Distributed Computing, pages 77--101, 1986.

....model Listeners because of its similarity with the Listeners object oriented pattern introduced by Gamma et al. 8] relying on a snapshot. As a result, SBQ L provides strong consistency semantics using fewer servers. In particular, Table 1 shows that SBQ L provides atomic semantics [9] for generic data using as few as 3f 1 servers to tolerate f faults, instead of the 4f 1 servers that were previously required to provide even the weaker regular [11] or partial atomic [16] semantics. We show that SBQ L is optimal with respect to the number of servers required to provide a ....

....attention to server failures and assume that clients are correct. We relax this assumption in Section 7.1. 2.2 Consistency Semantics Consistency semantics de ne system behavior in the presence of concurrency. We rst review Lamport s de nitions of safe, regular, and atomic semantics. Lamport [9] de nes the three semantics for distributed shared memory listed below. His original de nitions exclude concurrent writes, so we present extended de nitions that include these [16] Using a global clock, we assign a time to the start and end (or completion) of each operation. We say that an ....

L. Lamport. On interprocess communications. Distributed Computing, pages 77-101, 1986.

....write it, as well as the number of values it can take on. Registers are also classified according to the consistency guarantees they provide in the presence of concurrent operations. Three kinds of consistency guarantees, namely safeness, regularity and atomicity,have been defined by Lamport in [2] and have become of fundamental importance in the study of shared registers. According to those definitions, a shared register, which can be concurrently accessed by one writer process and one or more reading processes, is called: safe if it guarantees only that a read which does not happen ....

....dimensions imply a hierarchy on registers. The idea is to start with simple communication primitives (such as single writer single reader safe registers) which can be provided directly in hardware, and successively construct more powerful multi reader (even multi writer) multi valued objects [2, 3]. This procedure leads to modular system organization. 1.2. Contribution of this paper and related work Despite the fact that there has been a great deal of research on implementations of stronger registers out of weaker ones [2,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20] to the best of our ....

[Article contains additional citation context not shown here]

Lamport, L. (1986) On interprocess communication, part I: basic formalism, part II: basic algorithms. Distrib. Comput., 1, 77--101.

No context found.

Leslie Lamport. On interprocess communication. Distributed Computing, 1:77-101, 1986.

No context found.

Lamport, L. On interprocess communication. Distributed Computing 1 (

No context found.

Leslie Lamport. Interprocess Communication. Technical Report, SRI International, March 1985.

No context found.

Leslie Lamport. On interprocess communication. Distributed Computing, 1:77-101, 1986.

No context found.

Lamport L: On Interprocess Communications: I. Basic Formalism. Distributed Computing, 1: 77-85 (1986)

No context found.

Leslie Lamport, On Interprocess Communication,inDistributed Computing, vol.1,77--101. 12

No context found.

Leslie Lamport, "On Interprocess Communication", Distributed Computing 1 (1986), 77-- 101.

No context found.

L. Lamport. On interprocess communications. Distributed Computing, pages 77--101, 1986.

No context found.

L. Lamport. On interprocess communication. Distributed Computing, 1(2):77-101, 1986.

No context found.

Lamport, L.: On Interprocess Communication. In Distributed Computing 1,2 1986.

No context found.

L. Lamport. On interprocess communications. Distributed Computing, pages 77--101, 1986.

No context found.

L. Lamport. On interprocess communications. Distributed Computing, pages 77--101, 1986.

No context found.

L. Lamport, "On Interprocess Communication", Distributed Computing, 1 (1986), pp. 77-101.

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