Implementation of a portable software DSM in Java

Abstract

Rapid commoditization of advanced hardware and progress of networking technology is now making wide area high-performance computing a.k.a, the 'Grid' Computing a reality. Since a Grid will consist of vastly heterogeneous sets of compute nodes, especially commodity clusters, some have articulated the use of Java as a suitable technology to satisfy portability across different machines. Since Java's natural model of parallelism is shared memory multithreading, one will have to support distributed shared memory (DSM) in a portable manner; however, none of the previous work on implementing Java on DSM has been a portable solution. Instead, we propose a software architecture whose goal is to achieve portability of DSM implementations across different commodity clustering platforms, while restricting the programming model somewhat, and implemented a prototype system, JDSM. Benchmark resuits show that the current implementation on Java incurs increased memory coherency maintenance cost compared to C-based DSMs, thus limiting scalability to some degree, and we are currently working on a solution to alleviate this cost.

A Platform-Independent Distributed Runtime for Standard Multithreaded Java

“…Note that this configuration is not very advantageous in the context of distributed execution, since the ratio of bandwidth to CPU power is considerably smaller than in performance evaluations of similar systems. (1,4,19) The results obtained by this hardware setting when running on Red Hat Linux are not presented, due to their similarity to the results obtained under Windows XP. Finally, we compare the performance of a less scalable application, on a collection of Intel 2 × 2.4 MGHz dual processor machines running on Red Hat Linux in two communication settings: 100 Mbps Ethernet and 10 Gbps Inifiniband in IP-over-Infiniband (IPoIB) mode.…”

“…Among the systems that deviate from the Java programming paradigm, JavaParty (3) and JSDM (4) are close enough to pure Java in order to be considered in the current context. JavaParty supports parallel programming by extending Java with a preprocessor and a runtime.…”

“…Existing systems that are able to execute a Java program on top of a collection of standard JVMs (3−6) are not transparent. Some works, such as (3) and, (4) introduce unorthodox programming constructs. Others, such as (5) and, (6) require user intervention to enable distributed execution of existing programs.…”

Engineering Distributed Shared Memory Middleware for Java

“…For example, TreadMarks propagates updates to all nodes while our solution "hits" interested nodes only. JDSM [25] uses a centralized node (the Cluster Manager) to intercept requests, creating a bottleneck as the number of nodes increases.…”

Scaling a Shared Object Space to the Internet: Case Study of Virat.

“…Two other models have endeavoured to extend DSM across clusters: JDSM [31] and Lemuria [32]. JDSM presents an implementation of software DSM written in Java that can be ported to several cluster platforms.…”