ABSTRACT
In high-performance scientific computing (HPC), parallel sparse linear solvers are ubiquitous and the often are the most computationally intensive parts. Many existing packages with different APIs and languages are widely used in the HPC community. This prevents applications built using packages from changing easily to another package.The Common Component Architecture (CCA) framework provides the ability to encapsulate complex scientific software and to reuse it easily. Interfaces are written in scientific interface definition language (SIDL) and language interoperability can be achieved through the BABEL compiler from LLNL. CCA-LISI spans multiple widely-used HPC sparse linear solver packages and deploys a common API to capture the interaction between HPC applications and HPC solvers.However, interface designing in HPC raises new challenges. Some of issues and design decisions are explained, a proposed interface is shown, and a test implementation presented showing that the overhead introduced by CCA framework is minor.
Index Terms
- CCA-LISI: on designing a common component architecture parallel sparse linear solver interface
Recommendations
Merging the CCA Component Model with the OGSI Framework
CCGRID '03: Proceedings of the 3st International Symposium on Cluster Computing and the GridThe most important recent development in Grid systemsis the adoption of the Web services model as its basic architecture. The result is called the Open Grid ServicesArchitecture (OGSA). This paper describes a componentframework for distributed Grid ...
CCA: a deadline-constrained workflow scheduling algorithm for multicore resources on the cloud
Workflows are adopted as a powerful modeling technique to represent diverse applications in different scientific fields as a number of loosely coupled tasks. Given the unique features of cloud technology, the issue of cloud workflow scheduling is a ...
MOCCA - Towards a Distributed CCA Framework for Metacomputing
IPDPS '05: Proceedings of the 19th IEEE International Parallel and Distributed Processing Symposium (IPDPS'05) - Workshop 4 - Volume 05We describe the design and implementation of MOCCA, a distributed CCA framework implemented using the H2O metacomputing system. Motivated by the quest for appropriate metasystem programming models for large scale scientific applications, MOCCA combines ...
Comments