Albert I. Reuther & Joel Goodman HPEC 2003 25 Sept 2003 Dynamic Resource Management for a Sensor-Fusion Application via Distributed Parallel Grid Computing Albert I. Reuther & Joel Goodman HPEC 2003 25 Sept 2003 This is the introduction slide which includes the title, the authors, and sponsorship information. This work is sponsored by the Defense Advanced Research Projects Agency (DARPA) under Air Force contract F19628-00-C-002. Opinions, interpretations, conclusions and recommendations are those of the author and are not necessarily endorsed by the U.S. Government. 1 1

Operator Assisted Search and Integration System (OASIS) Problem: How does one build grid computing capability to enable real-time sensor data fusion? Real-time processing Streamed signal processing applications Dynamic resource management Guaranteed Quality-of-Service Failover fault tolerance Network Beowulf SGI Origin SGI O2 NRM This slide introduces the research problem of the study: how to build grid computing capability to enable real-time sensor data fusion. It introduces a generic computer network, the requirements to enable such a capability, and brief a specific sensor data fusion application, OASIS. Operator Assisted Search and Integration System (OASIS)

Solution: Graph-based Network Resource Manager TASK 1 (Stage 1) TASK 2 (Stage 2) TASK c-1 (Stage c-1) TASK c (Stage c) Determined most effective use of resources Launched application tasks on resources Ran multiple applications on same resources Demonstrated fault tolerant capabilities This slide explains the solution to building the infrastructure. It is to use a graph-based network resource manager (NRM). A specific network of computers, a conceptual NRM graph, and a specific OASIS NRM graph are shown. Also a list of accomplishments of the study are given.