B. Nitzberg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 52--60, August 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....In this environment, tuplespace provides a shared memory abstraction on top of a message passing, distributed memory machine. Because tuplespace is a logically shared memory, it encounters implementation issues similar to other distributed shared memories (DSM) 33, 35, 5, 27] Nitzberg and Lo [36] present a survey of the important issues of implementing DSMs and discuss how the developers of several DSMs faced these design choices. However, because of its special nature, a tuplespace implementation must address several additional issues [44] Combining our own implementation experiences ....

Bill Nitzberg and Virginia Lo. Distributed shared memory: A survey of issues and algorithms. Computer, 24(8):52--60, August 1991.

....and that these events coincide with consistency requirements. Therefore, relaxed consistency models can often be introduced without any or with only minor code changes. Relaxed consistency models have been extensively researched in the realm of Software Distributed Shared Memory (SW DSM) 13] [14], 15] and implemented in the form of relaxed consistency protocols [16] 17] While latter ones are not applicable or necessary in HW DSM scenarios, the concept of relaxed consistency models can be applied in a rather straightforward way. The result for the programmer is a secure, global memory ....

B. Nitzberg and V. LO, "Distributed Shared Memory: A Survey of Issues and Algorithms," IEEE Computer, pp. 52--59, Aug. 1991.

....data, or at least do not unknowingly access data that is stale according to the rules of the consistency model. Problems related to cache consistency have been widely studied in many other systems such as multiprocessor architectures [7] distributed file systems [22] distributed shared memory [23], and client server database systems [10] Depending on whether or not the server maintains the state of the client s cache, two invalidation strategies are used: the stateful server approach and the stateless server approach. In the stateful server approach, the server maintains the information ....

B. Nitzberg and V. Lo, "Distributed Shared Memory: A Survey of Issues and Algorithms," Computer, vol. 24, no. 8, Aug. 1991.

....large sized data. The first design gives the programmers more control whereas the second design provides more transparency. The benefits of both designs include improved efficiency, ease of construction, and ease of use. I. INTRODUCTION Message Passing (MP) and Distributed Shared Memory (DSM) [1], 2] 3] 4] 5] are the two most common approaches to programming on distributed memory systems. MP is scalable but difficult to use, and DSM is easy to use but unscalable. The reason behind the efficiency of MP is that it encourages the programmers to place a subcomputation on the node ....

....OS and hardware. Examples of kernellevel DSM systems are Gobelins [21] and Genesis [22] Most of the DSM systems are implemented at a user (software) level. User level DSM systems provide a shared address space for shared data on top of physically distributed memory using software support [1]. They rely on user level memory management techniques provided by the OS to detect accesses and updates to shared data at the level of pages objects. The software DSM system then applies a memory coherence protocol to provide an illusion of shared memory. Some well known user level threadoriented ....

B. Nitzberg and V. Lo, "Distributed shared memory: A survey of issues and algorithm," IEEE Computer, vol. 24, no. 8, pp. 52--60, Aug. 1991.

....receive the data. The final picture is that of a difficult system to program. In part to address the difficulty of programming parallel distributed systems, Shared Memory Models are being considered as an alternative to the traditional communication model in distributed architectural platforms [6, 8, 12, 13, 15, 17]. These models provide a set of shared addresses that the processes can use in the same way they access data locally stored, i.e. using the primitives data=read(address) and write(address,data) In general, memory models give a simpler view than the communication model in applications where the ....

Nitzberg, B., Lo, V. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 52-60, August 1991.

....larger communication delay of the DSM system makes the consistency protocol for the distributed memory slow and complex. The communication overhead to preserve the consistency is considered as a major cause of the performance bottleneck and makes the software DSM systems less scalable [3]. Several memory consistency protocols other than the standard sequential consistency [4] have been proposed [5, 6] to overcome the communication overhead and to increase the scalability degree of the DSM systems. Weak consistency [7, 8] processor consistency [9] and release consistency [10] ....

B. Nitzberg and V. Lo, "Distributed shared memory: A survey of issues and algorithms," in IEEE Computer, Vol. 24, No. 8, pages 52-60, Aug. 1991.

....the occurrence of an event. Hence, each communication step used for synchronization is dominated by t. This makes the overhead of a barrier synchronization proportional to the number of communication steps involved. As the distributed memory multiprocessors have a high communication start up cost [2], there is a strong need to develop efficient barrier synchronization schemes requiring fewer number of communication steps. Multicomputer networks are being used to build largescale distributed memory systems (Intel Paragon) 13] clusterconnected shared memory systems (Stanford DASH) 10] and ....

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. 1EEE Computer, pages 52 60, Aug. 99L

....implementations. The abstractions should in no way restrict the application programmer. Consider cooperation models for example. Several papers have emphasized the performance advantage of message based implementations of distributed applications, as compared to shared data implementations [Ni91]. However, our own experience with building distributed RBSOR [Co91] on ARCADE, using both message passing and DSM, disproves the universality of such claims. The DSM based implementation of RBSOR actually performed better, inspite of using a larger number of messages, because of the ability to ....

Nitzberg, B. and Lo, V., "Distributed Shared Memory: A Survey of Issues and Algorithms", IEEE Computer, Vol. 24, No. 8, August 1991, pp. 52-60.

....for multiple cooperating implementations and low enough to support a variety of computational models. 7. 2 Data Units and Distributed Shared Memory Distributed shared memory has been implemented in hardware [le90] as operating system software [Li86] and through compiler generated code [Ba89][Ni91]. The major DSM design issues are granularity of shared data, coherence protocol and support for heterogeneity [Ni91] Ivy [Li86] classically assumes shared data is totally unstructured, using hardware dependent page based granularity. Linda s shared data is a tuple space [Ah89] Ca90] defining ....

.... Units and Distributed Shared Memory Distributed shared memory has been implemented in hardware [le90] as operating system software [Li86] and through compiler generated code [Ba89] Ni91] The major DSM design issues are granularity of shared data, coherence protocol and support for heterogeneity [Ni91]. Ivy [Li86] classically assumes shared data is totally unstructured, using hardware dependent page based granularity. Linda s shared data is a tuple space [Ah89] Ca90] defining application dependant tuple29 based granularity. Munin [Ct91] structures its shared data on the basis of variables in ....

[Article contains additional citation context not shown here]

Nitzberg, B. and Lo, V., Distributed Shared Memory: A Survey of Issues and Algorithms, IEEE Computer, Vol. 24, No. 8, August 1991, pp. 52-60.

....new ways to do distributed computing. Distributing System V shared memory provides Linux with DSM capabilities. Processes on different machines can read the shared memory at the same time, and the effects of a write will automatically become visible to others. Under DIPC, strict consistency [6] is used as the consistency model of the shared memory. This means that the readers of the shared memory get the most recently written values. This is very familiar to programmers. Programs can use the shared memory as an asynchronous way of exchanging data, or they can use System V messages to ....

Bill Nitzberg and Virginia Lo, "Distributed Shared Memory: A survey of Issues and Algorithms", Computer, August 1991, pp. 52-60.

....implying that they can be read or written any time an application programmer wants to. There is no need to do any special synchronization activity before or after the access. DIPC s shared memory uses the single writer multiple readers protocol [7] and employs the strict consistency model [4], meaning that a read will return the most recently written values. Read or write access rights are assigned to computers, and all the processes on that machine can enjoy the rights. DIPC can manage shared memories in two modes: In the segment mode, whenever a process wants to read or write a ....

Bill Nitzberg and Virginia Lo, "Distributed Shared Memory: A Survey of Issues and Algorithms", Computer, 1991.

.... these are (i)availability of powerful processors with efficient hardware support for fast context switching, ii)availability of inexpensive memory so that each node can have enough local memory to support multiple processes, iii)recent development of the concept of distributed shared memory [28][29] that allows a shared memory programming environment in distributed memory systems and finally (iv)the concept of shared memory hypercubes [30] and meshes [11] that is becoming attractive to take advantage of the shared memory design on direct interconnection networks. Putting all of these ....

B.Nitzberg and V.Lo, "Distributed Shared Memory: A Survey of Issues and Algorithms, " IEEE Computer, Vol.24 pp.52-60, Aug.1991. 148

....of real writes, the evaluations of the variables last longer. As a result, after the learning period, ap saves 33 of the messages compared with kld. 5 Related work and conclusions Distributed shared memory provides a very interesting way to use distributed memory machines. The survey paper [NL91] introduces all the notions of granularity, coherency semantics and coherence protocols inherent to distributed shared memory systems. Several systems have been built on the algorithms proposed by Li and Appel [LH89] IVY [Li86] CHORUS [ARG89] KOAN [LP91] etc. A lot of work have been done in ....

Bill Nitzberg and Virginia Lo. Distributed shared memory: a survey of issues and algorithms. Computer, 24(8):52--60, August 1991. 15

....the distributed system. Also, normal Linux kernels are not affected by DIPC programs, meaning that the there is no need to change and recompile these programs when they are to be executed in single computers with no DIPC support. DIPC s Distributed Shared Memory Distributed Shared Memory (DSM) [6] in DIPC uses a multiple readers single writer protocol: DIPC replicates the contents of the shared memory in each computer with reader processes, so they can work in parallel, but there can be only one computer with processes that write to a shared memory. The strict consistency model is used ....

Bill Nitzberg and Virginia Lo, "Distributed Shared Memory: A Survey of Issues and Algorithms", Computer, August 1991, pp 52-60.

.... future distributed systems [3] Replication of objects in a distributed system has been used to improve performance and also to provide fault tolerance in different application areas such as: persistent data management [2] management of replicated computations [15] and distributed shared memory [14]. These three application areas can be considered complementary and none of them fully covers the spectrum of applications that can use replicated data management. For instance, while persistent data management systems often preclude the efficient sharing of volatile objects, distributed ....

Bill Nitzberg and Virginia Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 24(8):52--60, August 1991.

....computing, image processing, simulation and optimization. 1. The DIPP programming model Ideally, explicit parallel programming should be uniform for shared and distributed platforms. One way to achieve this is to give the programmer the illusion of shared memory even on distributed platforms[18]. While we think that this might be exactly what a programmer wants we must acknowledge that this goal is not reached yet. The performance issue implies some sort of discrimination of local from non local data for distributed platforms. We believe that the process model can be uniform without any ....

B. Nitzberg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE COMPUTER, vol. 24, no. 8, pp. 52-60, Aug 1991.

....has become increasingly important to resolve. While the traditional approaches are based on message passing, the more comfortable shared memory model, previously restricted to tightly coupled system like SMPs, is gaining acceptance in the form of Distributed Shared Memory (DSM) systems [18]. They provide a single, global address space, allow inherent data distribution, and support incremental parallelization and therefore a smooth migration path for the parallelization of an application. This ease of use, however, comes at a price. Due to the transparency provided by the shared ....

B. Nitzberg and V. LO. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, pages 52--59, Aug. 1991.

....model for shared memory programming. Moreover, the process models are often interwoven with the data model. Some of the well known programming models are designed for one of the two memory models 1 If we talk about shared memory, this shall also include distributed shared memory (see e.g. [15]) only. PVM, for instance, has been developed for distributed programming. It does not support shared memory programming. Since its main concept, message passing is defined in terms of process communication there is no separation of process and data models. Few programming models support logical ....

B. Nitzberg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE COMPUTER, vol. 24, no. 8, pp. 52-60, Aug 1991.

....and conclude with implementations issues and benchmarks. 2 Introduction and related work Ideally, explicit parallel programming should be uniform for shared and distributed platforms. One way to achieve this is to give the programmer the illusion of shared memory even on 2 distributed platforms[NL91] While we think that this might be exactly what a programmer wants we must acknowledge that this goal has not been reached yet. The performance issue implies some sort of discrimination of local from non local data for distributed platforms. This may be done at the level of the objects a ....

Bill Nitzberg and Virginia Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE COMPUTER, vol. 24, no. 8, pp. 52-60, Aug 1991.

....in an algorithm, rather than the concrete parallelism of a machine. For communication, Filaments provides shared variable programming, because it is simpler than other communication models. Moreover, it is the mostly widely used model, as most algorithms are expressed in terms of shared memory [46]. For synchronization, Filaments primarily provides barrier synchronization because it is simple and sufficient. 2 A barrier is a logical fence at which all processes must arrive before any may pass. Filaments supports parallel scientific applications, which are characterized by a large amount ....

....to a program, such as flow control and buffering of messages. Programming with messages is tricky despite the various solutions that have been proposed [52] Moreover, the shared variable model is the most widely used model; for example, most algorithms are expressed in terms of shared memory [46]. Figure 3.2 shows how Jacobi iteration might be programmed using explicit mes 49 do maxdiff epsilon # send row up and receive from below if not firstNode send (above, new[firstRow] rowsize) fi if not lastNode receive (below, new[lastRow 1] rowsize) fi # send row down and receive ....

Bill Nitzberg and Virginia Lo. Distributed shared memory: A survey of issues and algorithms. Technical Report CIS-TR-90-26, University of Oregon, January 1991.

..... 12 4 Performance evaluation 13 5 Conclusion 16 LIP, ENS Lyon, 46 Alle d Italie, F 69364 Lyon Cedex 07, France. Contact: fGabriel. Antoniu, LucBougeg ens lyon.fr. 1 1 Introduction In their traditional flavor, Distributed Shared Memory (DSM) libraries [16, 19, 20, 11] allow a number of separate processes to share a common address space using a consistency protocol according to a semantics specified by some given consistency model: sequential consistency, release consistency, etc. The processes may usually be physically distributed among a number of computing ....

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE computer, 24(8):52--60, September 1991. 18

....overview 3 3 Specifying protocols in DSM PM2 5 4 Performance evaluation 7 5 Conclusion 10 LIP, ENS Lyon, 46 Alle d Italie, F 69364 Lyon Cedex 07, France. Contact: fGabriel. Antoniu, LucBougeg ens lyon.fr. 1 1 Introduction In their traditional flavor, Distributed Shared Memory (DSM) libraries [7, 10, 11, 4] allow a number of separate processes to share a common address space according to some fixed consistency model: sequential consistency, release consistency, etc. The processes may usually be physically distributed among a number of computing nodes interconnected by some communication library. The ....

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE computer, 24(8):52--60, September 1991.

....clusters. We illustrate its power with figures obtained for different protocols implementing sequential consistency, release consistency and Java consistency, on top of Myrinet, Fast Ethernet and SCI clusters. 1 Introduction In their traditional flavor, Distributed Shared Memory (DSM) libraries [7, 10, 11, 4] allow a number of separate processes to share a common address space according to some fixed consistency model: sequential consistency, release consistency, etc. The processes may usually be physically distributed among a number of computing nodes interconnected by some communication library. The ....

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE computer, 24(8):52--60, September 1991.

....the AM, and discuss the trade off between additional unallocated memory space and higher set associativity. 1. Introduction Distributed shared memory multiprocessors with non uniform memory architecture (NUMA) have gained approval in recent years because of their potential for higher scalability [11,12]. Careful placement, replication, and migration of data can ensure that the majority of memory references by each processor is to the fraction of memory address space phy sically local to its own processing node [5,16] Unlike traditional NUMA, the cache like memory in each processing node, called ....

Bill Nitzberg and Virginia Lo, "Distributed Shared Memory : A survey of Issues and Algorithms", Computer, Aug., 1991, pp 52-60.

....an idle node will poll other nodes for work at set intervals. Overloaded nodes can then send filaments to the idle node. 2 2.2.3 Distributed Shared Memory Filaments applications share information via a distributed implementation of shared memory. The goal of distributed shared memory(DSM) [NL91] is to provide an application running on a distributedmemory machine with the illusion of a shared memory space. Distributed shared memory (DSM) takes separate memory space from several different machines and treats it as though the same memory existed on all machines. It works in much the same ....

Bill Nitzberg and Virginia Lo. Distributed shared memory: A survey of issues and algorithms. IEEE Computer, 24:52--60, August 1991. 10

....shared virtual memory in [19] as an extension to the operating system s virtual memory management, a lot of research has been done to improve the performance of virtual shared memory systems 2 . The problems which have to be solved are the high latency for remote data accesses and false sharing [26]. In order to integrate virtual shared memory management in the operating system s virtual memory manager many systems use virtual memory pages as the unit of transportation and coherency. If the granularity of these units does not fit with the way the program accesses the shared data, the page ....

....based virtual shared memory system Munin [2] as in the Dash system, release consistency [14] is used to do this. Munin additionally provides different types of shared variables with strict or weak coherency protocols. An overview of virtual shared memory algorithms and systems can be found in [26], 30] 32] and [33] In Pact, virtual shared memory management is done with a fixed distributed manager algorithm with write invalidation similar to Li s algorithm in [20] This management is done in the runtime system instead of the operating system. This has the advantage that Pact can be ....

[Article contains additional citation context not shown here]

Bill Nitzberg and Virginia Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 24(8):52--60, August 1991.

....by the INRIA ResCapA Research Coordinated Action, the NSF INRIA C IT Cooperative Research Grant, the CNRS ARP Research Program and the ReMaP Project, INRIA Rhne Alpes. 1 1 Introduction Distributed Shared Memory (DSM) libraries have been available for a dozen of years. In their traditional avor [7, 11, 12], they allow a number of separate processes to share a common address space according to some xed consistency model: sequential consistency, release consistency, etc. The processes may usually be physically distributed among a number of distant computing nodes interconnected by some communication ....

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE computer, 24(8):5260, September 1991. 12

....Some people point out similarities between Planet s basic functions and those of other operatingsystem or middleware level memory management systems. These include distributed shared memory (DSM) and single address space (SAS) systems. Planet s mobile memory segment is related to DSM [29, 22] and SAS systems like Opal [9] in that memory areas for application programs are transferred between computer sites in an automatic, transparent way. The difference is in their address sharing methods. DSM and SAS provide programmers with a virtual view wherein single address space may exist. ....

B. Nitzberg and V. Lo. Distributed shared memory: a survey of issues and algorithms. IEEE Computer, 24(8):52--60, Aug. 1991.

....accesses. Applications running on DSM systems are invariable slower than those run on shared memory multiprocessors. However, they should show a speedup over sequential execution to demonstrate their usefulness. Scalable. One of the benefits of DSM systems mentioned in much of the literature [Nitzberg and Lo 94] Tanenbaum 95] is that they scale better than many tightlycoupled shared memory multiprocessors. However, scalability is limited by physical bottlenecks, e.g. operations which require global information or distribute information to all workstations in the COW (broadcast messages) Nitzberg and ....

....and Lo 94] Tanenbaum 95] is that they scale better than many tightlycoupled shared memory multiprocessors. However, scalability is limited by physical bottlenecks, e.g. operations which require global information or distribute information to all workstations in the COW (broadcast messages) Nitzberg and Lo 94] There should be no potential bottlenecks to limit the scalability of the system. We have decided to provide an update based DSM on RHODOS in parallel to the existing invalidation based DSM both of which adhere to the above design requirements. The inclusion of an update based DSM system will ....

[Article contains additional citation context not shown here]

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994, pp 375-386.

....form of caching which is often used to improve performance in DSM systems. However, the maintenance of the coherence of this replicated memory in a transparent and efficient way is an important research objective because computer network latency is typically much larger than that of a shared bus [Nitzberg et al. 94] Tam et al. 90] 28 May 1997 2 The advantages of DSM [Coulouris et al. 93] Nitzberg et al. 94] Levelt et al. 92] are that it: Increases the ease of programming by sparing programmers the concerns of message passing; Allows the use of algorithms and software written for shared ....

.... of the coherence of this replicated memory in a transparent and efficient way is an important research objective because computer network latency is typically much larger than that of a shared bus [Nitzberg et al. 94] Tam et al. 90] 28 May 1997 2 The advantages of DSM [Coulouris et al. 93] Nitzberg et al. 94] Levelt et al. 92] are that it: Increases the ease of programming by sparing programmers the concerns of message passing; Allows the use of algorithms and software written for shared memory multiprocessors on distributed systems; Distributed machines scale to much larger numbers ....

[Article contains additional citation context not shown here]

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994, pp 375-386.

....of non local memory should be as close to the access time of local memory as possible. This is also related to transparency in that the programmer should see no discernable difference between memory accesses. Scalable. One of the benefits of DSM systems mentioned in much of the literature [Nitzberg et al., 94] Tanenbaum, 95] is that they scale better than many tightly coupled shared memory multiprocessors. However, scalability is limited by physical bottlenecks, e.g. operations which require global information or distribute information globally (broadcast messages) Nitzberg et al., 94] There ....

....literature [Nitzberg et al., 94] Tanenbaum, 95] is that they scale better than many tightly coupled shared memory multiprocessors. However, scalability is limited by physical bottlenecks, e.g. operations which require global information or distribute information globally (broadcast messages) Nitzberg et al., 94] There should be no potential bottlenecks to limit the scalability of the system. To achieve these design requirements the following issues must be considered. 28 May 1997 3 2.1.1 Granularity The granularity of the blocks of data to be moved around the distributed system can be one of the ....

[Article contains additional citation context not shown here]

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994, pp 375-386.

....is twice the message delay. In recent years, there has been wide spread interest in implementing shared memory programming paradigm on distributed systems. Such systems, called Distributed Shared Memory (DSM) make it possible to execute shared memory parallel programs on a distributed system [8, 9, 10, 11, 12, 13].This makes it all the more important to provide shared variable like synchronization mechanisms in distributed systems. Most DSM systems provide centralized implementations of mechanisms such as semaphores. A notable exception is Munin, which provides distributed locks for synchronization [14] ....

Bill Nitzberg and Virginia Lo. "Distributed Shared Memory: A Survey of Issues and Algorithms". IEEE Computer, pages 52--60, August 1991.

....shared memory architecture. Before building the actual computer it is desirable to be able to make some kind of evaluation of the architecture s performance. The easiest way to do this is to use a simulator. 1. 1 A distributed shared memory architecture The distributed shared memory architecture [Nitzberg 91] is a mixture of the two most common architectures for parallel computers, distributed memory architecture and shared memory architecture, and is predicted to give a better performance than these traditionally used architectures. To see how a distributed shared memory architecture works, we first ....

Bill Nitzberg, Virginia Lo. Distributed Shared Memory: A Survey of Issues and Algorithms, IEEE Computer, Aug 1991, pp. 52-60.

.... processes) 93, 94] In most systems the address space and threads are restricted to reside on a single node (a uni or multi processor) However, some systems such as Ivy, the Apollo domain, and Clouds [33] support a distributed address space (sometimes called a distributed shared memory see [82] for a summary of issues involved with distributed shared memory) and distributed threads executing on that address space. Regardlesss of whether the address space is local or distributed [42] there has been significant work done the following topics. They are: supporting very large, but sparse ....

B. Nitzberg and V. Lo, "Distributed Shared Memory: A Survey of Issues and Algorithms, " IEEE Computer, Vol. 24, No. 8, August 1991, pp. 52-60. 35

....sections. The reason is that the simulator cannot enforce the correct order of access to shared variables outside critical sections since the processes can access shared variables without stopping their execution. This restriction is in accordance with a weakly consistent shared memory semantics [Nitzberrg Lo 91] 4 Execution Driven Simulation with SPAM The SPAM kernel is made up of two parts: a tracing tool and a simulation library. The tracing tool generates the address traces and synchronization events of the processes of the parallel application under simulation. The simulation library allows the ....

B. Nitzberrg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 24(8):52--60, August 1991.

....protocols (BIP, SCI, VIA, MPI, TCP, etc. We report performance gures for three platforms using di erent network protocols: SISCI SCI, BIP Myrinet and TCP Myrinet. 1 Introduction Distributed Shared Memory (DSM) libraries have been available for a dozen years. In their traditional avor [9, 14, 1, 15], they allow a number of processes to share a common memory space according to some xed consistency model: strong consistency, release consistency, etc. The processes may often be distributed among a number of computing nodes interconnected by some communication library. The design of the library ....

B. Nitzberg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE computer, 24(8):5260, September 1991.

No context found.

B. Nitzberg and V. Lo. Distributed Shared Memory: A Survey of Issues and Algorithms. IEEE Computer, 52--60, August 1991.

No context found.

B. Nizeberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE Computer, vol. 24, no.8, pp. 52-60, Aug. 1991.

No context found.

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE Computer, 24(8):52-60, August 1991.

No context found.

B. Nitzberg, V. Lo, Distributed shared memory: a survey of issues and algorithms, IEEE Comput., August 1991, pp. 52--59.

No context found.

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE Computer, pages 52--60, 1991.

No context found.

B. Nitzberg and V. Lo, `Distributed shared memory: a survey of issues and algorithms', IEEE Computer, 24, (8), 52--60 (1991).

No context found.

B. Nitzberg, V. Lo, Distributed Shared Memory: A survey of Issues and Algorithms, Computer, vol. 24, n 8, Aug. 90, pp. 5-60

No context found.

Nitzberg, B. and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE Computer, 24(8):52-60, August 1991.

No context found.

B. Nitzberg and V. Lo. Distributed shared memory: a survey of issues and algorithms. Computer, 24(8):52-60, Aug 1991.

No context found.

Nitzberg, B. and V. Lo. 1991. "Distributed Shared Memory: A Survey of Issues and Algorithms", Computer, 24, 52-60.

No context found.

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994, pp 375-386.

No context found.

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994.

No context found.

B. Nitzberg, V. Lo, Distributed Shared Memory: A Survey of Issues and Algorithms, In Casavant T.L. and Singal M. (eds), Readings in Distributed Computing Systems, IEEE Press, 1994, pp 375-386.

No context found.

B. Nitzberg and V. Lo. Distributed shared memory: A survey of issues and algorithms. IEEE computer, 24(8):5260, September 1991. 9

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