Accession Number : ADA290133

Title :   Fault Tolerant Features and Experiments of ANTS Distributed Real-Time System. ANTS: An Approach for High-Performance and Ultra-Dependability.

Descriptive Note : Semiannual, 16 Jul 94-15 Jan 95,

Corporate Author : RHODE ISLAND UNIV KINGSTON

Personal Author(s) : Dominic-Savio, Patrick ; Lo, Jien-Chung ; Tufts, Donald W.

PDF Url : ADA290133

Report Date : 15 JAN 1995

Pagination or Media Count : 41

Abstract : The ANTS project at the University of Rhode Island introduces the concept of Active Nodal Task Seeking (ANTS) as a way to efficiently design and implement dependable, high-performance, distributed computing. This paper presents the fault tolerant design features that have been incorporated in the ANTS experimental system implementation. The results of performance evaluations and fault injection experiments are reported. The fault-tolerant version of ANTS categorizes all computing nodes into three groups. They are: the up-and- running green group, the self-diagnosing yellow group and the failed red group. Each available computing node will be placed in the yellow group periodically for a routine diagnosis. In addition, for long-life missions, ANTS uses a monitoring scheme to identify faulty computing nodes. In this monitoring scheme, the communication pattern of each computing node is monitored by two other nodes.

Descriptors :   *DISTRIBUTED DATA PROCESSING, *FAULT TOLERANT COMPUTING, TEST AND EVALUATION, MONITORING, REAL TIME, PROTOTYPES, NODES, MISSIONS, PATTERNS, DIAGNOSIS(GENERAL), OPERATING SYSTEMS(COMPUTERS), LONG LIFE, COMMUNICATION AND RADIO SYSTEMS, FAULTS, RHODE ISLAND.

Subject Categories : Computer Programming and Software

Distribution Statement : APPROVED FOR PUBLIC RELEASE