Grid Computing Research Laboratory State University of New York (SUNY) Binghamton Department of Computer Science

Rainer Schmidt, Michael R. Head, Madhusudhan Govindaraju, Michael J. Lewis, and Siegfried Benkner,
"Design and Implementation Choices for Implementing Distributed CCA Frameworks,"
GECO-COMPFRAME06: Workshop HPC Grid programming Environments and COmponents and Component and Framework Technology in High-Performance and Scientific Computing (at HPDC-15)
pp. 3--10, Paris, France, June 2006
[PDF] [PS] [bibtex]

Abstract
The Common Component Architecture (CCA) specification is designed to provide a plug-and-play environment for scientists to manage the complexity of large-scale scientific simulations. The same specification is used for the implementation of sequential, parallel and distributed frameworks. The CCA specification places minimal requirements on the framework design, thus allowing various research groups to manage the complexity of the underlying run-time systems in ways that match the performance requirements of their target applications. In this paper we discuss the various design choices, constraints and complexities of implementing the CCA specification for high-performance distributed applications. In particular, we focus on the following CCA features: component instantiation and connections, port type representations, Builder Service design, choice of middleware, remote component communication, registration and discovery, client interface and QoS. We present a discussion on the the design space of distributed CCA frameworks with specific examples from three concrete implementations: VGE-CCA, XCAT-C++ and LegionCCA.

Key Words: