P. Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, Mar 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....runtime speedup. In addition to the performance benefit, adaptive consistency protocols can also alleviate an application programmer of the burden to analyze an application to determine and specify the best protocol. We now discuss these works. 14 2.2. 1 CVM and Adaptive Protocols Keleher [Kel98, Kel99a] modified CVM to provide a access pattern based invalidate update hybrid protocol. The goal is to reduce fault latency for all shared data that a processor has accessed, under the pretense the processor will access that data in the future. Fault latency is the amount of time a processor ....

....chapter discusses conventional approaches to select the best consistency protocol for an application based on analysis of the application s shared data access patterns. Recently, a few DSM systems proposed using information about the access pattern to select the best protocol [MB98, ACD 99, Kel98, Kel99a] These systems have shown that this approach can lead to significant runtime speedups. However, these systems have several limitations. First, they are designed for LAN environments and do not consider issues that exist in WAN environments. For example, they do not consider how a ....

[Article contains additional citation context not shown here]

Pete Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, Mar 1998.

....because we perform eager update (at the time of a release) and we base our predictions on patterns of access requests rather than just a single prior access by the remote node. Keleher has also proposed a barrier only speculative protocol for applications with extremely regular access patterns [10]. Our protocol supports a wider range of synchronization mechanisms and more general access patterns. Finally, the HLRC on VIA system [17] also has as its goal reducing latency of remote accesses, but by using improved communication mechanisms. The Virtual Interface Architecture (VIA) is ....

P. Keleher. Update protocols and iterative scientific applications. In The 12th International Parallel Processing Symposium (IPPS), March 1998.

....fault nor requires any summary of write modifications, ii) it takes only one round trip message to bring a remote copy of any shared memory page up to date, and (iii) no garbage collection is needed. Due to these advantages, the home based SDSM protocol has been a focus of several recent studies [7, 10, 14, 18]. Unfortunately, no prior work has ever been attempted on crash recovery in home based SDSM. This paper is the very first one to deal with crash recovery in such SDSM. Message logging is a popular technique for providing home less SDSM with fault tolerant capability [6, 11, 12, 13, 17] This ....

P. J. Keleher. Update Protocols and Iterative Scientific Applications. In Proc. of the 12th Int'l Parallel Processing Symp. (IPPS'98), pages 675--681, March 1998.

.... KY 40506 USA fdiaz,griffg dcs.uky.edu Abstract Distributed Shared Memory (DSM) systems often support exactly one consistency model for all shared data [ACD 96, BZS93] Some recent systems support adaptive consistency using heuristical analysis of recent access patterns [ACD 99, Kel98, Kel99a, Kel99b, MB98] Although approaches based on access patterns can significantly improve an application s performance, there are other factors besides access patterns that influence the performance of a consistency model, such as network bandwidth, congestion, latency and topology. These ....

....all environments. Unfortunately, an application is often stuck with whatever consistency model the DSM provides. Recent work has explored the concept of adaptive consistency in which the DSM observes the application s access patterns and then selects an appropriate consistency model [ACD 99, Kel98, Kel99a, Kel99b, MB98] While access patterns provide some hints about which consistency model is the best, they do not necessarily tell the whole story [DG00] Other factors such as the size of data writes and elements related to the computing environment, such as congested links, heavily loaded ....

[Article contains additional citation context not shown here]

Pete Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, Mar 1998.

....Distributed Shared Data Christopher Diaz and James Griffioen Department of Computer Science University of Kentucky Lexington, KY 40506 USA fdiaz,griffg dcs.uky.edu Abstract. Distributed Shared Memory (DSM) systems typically support one consistency protocol [3, 5, 6] However, recent work [1, 11, 12, 14, 17] proposes the use of adaptive consistency based on a heuristical analysis of recent access patterns. Although heuristic based approaches can significantly improve runtime, the access pattern alone does not necessarily define the most appropriate consistency protocol. The size of updates and ....

....of factors. In most systems, the application is stuck with whatever consistency protocol the DSM provides. Some recent work, however, has explored the concept of adaptive consistency in which the DSM observes the application s access pattern and then selects an appropriate consistency protocol [1, 11, 12, 14, 17]. Given an application s recent access pattern, the DSM uses heuristics to map the observed access pattern to a consistency protocol. For example, if a machine writes to shared data that is subsequently read by N other machines, the DSM may decide to use an updatebased approach, such as that used ....

[Article contains additional citation context not shown here]

Pete Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, March 1998.

.... Data Technical Report CS300 00 Christopher Diaz James Griffioen Department of Computer Science University of Kentucky Lexington, KY 40506 USA fdiaz,griffg dcs.uky.edu Abstract Distributed Shared Memory (DSM) systems typically support one consistency model [3, 5, 6] However, recent work [1, 11, 12, 14, 16] proposes the use of adaptive consistency based on a heuristical analysis of recent access patterns. Although heuristic based approaches can significantly improve runtime, the access pattern alone does not necessarily define the most appropriate consistency model. The size of updates and other ....

....range of factors. In most systems, the application is stuck with whatever consistency model the DSM provides. Some recent work, however, has explored the concept of adaptive consistency in which the DSM observes the application s access pattern and then selects an appropriate consistency model [1, 11, 12, 14, 16]. Given an application s recent access pattern, the DSM uses heuristics to map the observed access pattern to a consistency model. For example, if a machine writes to shared data that is subsequently read by N other machines, the DSM may decide to use an updatebased approach (such as Eager Release ....

[Article contains additional citation context not shown here]

Pete Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, Mar 1998.

....and synchronization overhead due to high communication costs. With Prof. Keleher at Maryland, I investigated issues in combining compilers and software DSMs [23] To experimentally evaluate performance issues, I built COSMIC [41] a prototype compiler based on SUIF targeting the CVM software DSM [22]. We found two main methods for improving performance in software DSM systems: compiler directed coherence protocols and compile time synchronization elimination. Software DSMs typically use a lazy invalidate protocol to maintain coherence in shared memory. We found compilers can identify ....

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, April 1998.

....interprocessor messages (e.g. Fortran D [20] is to rely on sophisticated run time systems (e.g. CHAOS [5] PILAR [26] which can identify and gather nonlocal data. A second approach is to combine shared memory compilers (e.g. SUIF [11] with software DSM systems (e.g. TreadMarks [30] CVM [24]) which provide a shared memory interface. Software DSMs are less efficient than explicit messages, but are much simpler compilation targets [4,25] In this paper, we examine existing approaches to parallelizing irregular reductions, and propose a new efficient algorithm. 1.1 Irregular ....

....a number of techniques for improving efficiency. One approach relies on precise communication analysis to insert explicit messages when analysis exactly identifies interprocessor communication [3,4,7] A second approach exploits customized coherence protocols for reductions and nonlocal updates [8,24,25,33,39]. In addition, compilers can also eliminate unnecessary synchronization based on communication analysis [17,25] TreadMarks is one of the most efficient software DSMs currently available. It relies on an invalidate coherence protocol, and parallelizes irregular reductions using ReplicateBufs. ....

[Article contains additional citation context not shown here]

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, April 1998.

....data object but user defined parts of structured data objects derived from the sharing patterns of the parallel application are held consistent between the participating nodes. This allows to avoid false sharing for many applications. We applied a home based release consistent protocol [3, 4, 5] to reduce the communication overhead that is inherent to software DSM systems. The protocol was originally developed for page based systems but has been adapted to structured data objects instead of pages. Applied to the Rthreads system, the protocol is able to use the introduced description of ....

Pete Keleher. Update protocols and iterative scientific applications. In Proc. of the 12th Int. Parallel Processing Symp. (IPPS'98), pages 675--681, 1998.

....from IBM and DEC. is to rely on sophisticated run time systems (e.g. CHAOS [4] PILAR [15] which can identify and gather nonlocal data. A second approach is to combine shared memory compilers (e.g. SUIF [6] with software distributed sharedmemory (DSM) systems (e.g. TreadMarks [17] CVM [13]) which provide a shared memory interface. Software DSMs are less efficient than explicit messages, but are much simpler compilation targets [3, 14] In this paper, we introduce LocalWrite, a new compiler and run time parallelization technique which can improve performance for certain classes of ....

....We evaluate the performance of different parallelization approaches as we vary application characteristics, in order to identify areas in which software DSMs can match or even exceed the efficiency of explicit messages. Experiments are conducted in a prototype system [7, 14] using the CVM [13] software distributed shared memory (DSM) as a compilation target for the SUIF [6] shared memory compiler. Our paper makes the following contributions: develop and evaluate LocalWrite, a new compiler and run time technique for parallelizing irregular reductions based on the owner computes ....

[Article contains additional citation context not shown here]

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, April 1998.

....interprocessor messages (e.g. Fortran D [12] is to rely on sophisticated run time systems (e.g. CHAOS [4] PILAR [17] whichcan identify and gather nonlocal data. A second approach is to combine shared memory compilers (e.g. SUIF [7] with software DSM systems (e.g. TreadMarks [19] CVM [15]) which provide ashared memory interface. Software DSMs are less efficient than explicit messages, but are much simpler compilation targets [3, 16] In this paper, we introduce LOCALWRITE, a new compiler and run time parallelization technique which can improve performance for certain classes of ....

....We evaluate the performance of different parallelization approaches as we vary application characteristics, in order to identify areas in which software DSMs can match or even exceed the efficiency of explicit messages. Experiments are conducted in a prototype system [8, 16] using the CVM [15] software distributed shared memory (DSM) This research was supported by NSF CAREER Development Award #ASC9625531 in New Technologies. The IBM SP 2 and DEC Alpha Cluster were provided by NSF CISE Institutional Infrastructure Award #CDA9401151 and grants from IBM and DEC. as a compilation target ....

[Article contains additional citation context not shown here]

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, Apr. 1998.

....D [13] HPF [21] is to rely on sophisticated run time systems (e.g. CHAOS [5] PILAR [22] which can identify and gather nonlocal data. A second approach is to combine shared memory compilers (e.g. SUIF [8] with software distributed shared memory (DSM) systems (e.g. TreadMarks [24] CVM [18]) which provide a shared memory interface. Software DSMs are less efficient than explicit messages, but are much simpler compilation targets [4, 20] In this paper, we introduce LOCALWRITE, a new compiler and run time parallelization technique which can improve performance for certain classes of ....

....processors. Instead, the compiler relies on the observation that most scientific applications perform iterative computations. Compilers can thus utilize access histories from previous time steps to guide communication and aggregate messages to the same processor, reducing latency and overhead [18, 20, 30]. In addition, compilers can also eliminate unnecessary synchronization based on communication analysis [10, 20] For both approaches, experiments show performance is comparable to message passing codes for regular applications, and better than current commercial distributedmemory compilers for ....

[Article contains additional citation context not shown here]

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, April 1998.

No context found.

P. Keleher. Update protocols and iterative scientific applications. The 12th International Parallel Processing Symposium, Mar 1998.

No context found.

P. Keleher. Update protocols and iterative scientific applications. In Proc. of the 12th Int'l Parallel Processing Symposium (IPPS'98), pages 675 -- 681, 1998.

No context found.

P. Keleher. Update protocols and iterative scientific applications. In Proceedings of the 12th International Parallel Processing Symposium, Orlando, FL, April 1998.

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