Last modified 9 years ago Last modified on 03/25/10 13:43:54


Computational Quality of Service (CQoS)

To at least partially automate the process of characterizing the performance of component applications and selecting and configuring particular implementations, we have introduced the concept of computational quality of service for scientific applications. CQoS expands on traditional QoS ideas by considering additional application-specific metrics, referred to as metadata, that enable the characterization of the performance of high-performance components. Before automating the selection of component instances, however, one must be able to collect and analyze performance information and the related metadata.

We are developing a number of software tools to support this automation, including a C++ library for managing performance data and related metadata stored in an SQL database, as well as SIDL-based component interfaces for database management and performance data analysis.

1. Database Library (C++)

2. CQoS Multilanguage Interfaces and Components

3. Applications

  • PETSc example: 2-D driven cavity flow simulation
    • Prerequisites: PETSc, Anamod 1.5

  • Parallel mesh partitioning
  • Quantum chemistry