Introduction
Users
Documentation
Developers
Availability
Download


Applications
CUMULVS

Introduction

CUMULVS (Collaborative User Migration, User Library for Visualization and Steering) is a software framework that enables programmers to incorporate fault-tolerance, interactive visualization and computational steering into existing parallel programs. The CUMULVS software consists of two libraries, one for the application program, and one for the visualization and steering front-end (called the "viewer").

CUMULVS handles collecting and transferring distributed data fields to the viewers and oversees adjustments to steering parameters in the application. It also manages the dynamic attachment and detachment of multiple independent viewers to a running parallel application. In addition, CUMULVS provides a user-directed checkpoint/restart mechanism to enable users to integrate fault tolerance to a running parallel application.

This is a summary of the services that CUMULVS provides:
  • Collaborative on-line visualization of data produced by applications running locally or remotely. Data visualization may include multiple remote viewers.
  • Computational steering among several computer applications.
  • Fault-tolerance through an automatic recovery of the virtual environment to host or network failures.
  • Fault-tolerance of distributed computer applications through heterogeneous task migration and user-directed checkpointing.
  • System Monitoring for High-Performance.
  • Dynamic attachment and detachment of viewer.
  • Message-passing over wide-area ATM networks between multiple MPP and SMP systems.
  • Secure and authenticated data transfer.

Users

Any computer application needing to implement one or more of the following services can benefit from the use of CUMULVS:

  • Synchronization, coordination, and control inside parallel computations
  • Data Movement in a distributed data organization (considering data locality and latency hiding)
  • Monitoring and fault recovery through checkpoints and restart mechanisms
  • Data visualization from either massive storage or on-the-fly. This may include multiple viewers connected to an application running locally or remotely.
To instrument computer applications with CUMULVS, the users will only have to include a few simple calls to the CUMULVS libraries to initialize the CUMULVS environment, define and declare the fields to be visualized, and define steering parameters.

For instance, a MFE (Magnetic Fusion Energy) group at LLNL developed a code to visualize their parallel application using CUMULVS and AVS. Also, CUMULVS has been integrated into several DOE Computational Grand Challenges as well as DOE 2000 applications. CUMULVS was also used inside the CCAFFEINE (Common Component Architecture [CCA] Fast Framework Example In Need of Everything) solutions to several fundamental problems in the application of component software. The images in the left panel come from applications that benefited from the use of CUMULVS. Click on the small icons to display the full image (a new window will open).

Documentation

Developers

CUMULVS has been developed at the Computer Sciences Group at Oak Ridge National Laboratory. Its principal developers are James Kohl and Philip Papadopoulos.



Tools News Home