PROCESS CHECKPOINTING IN AN OPEN DISTRIBUTED ENVIRONMENT
Taha Osman, Andrzej Bargieła · 2011
This paper presents a fault tolerant environment (FTE) for the reliable execution of application programs on a network of workstations. The FTE consists of an integrated error detection and fault recovery system that efficiently detects processor node crashes and hardware transient failures, and recovers user-processes from these faults using checkpointing and roll-back techniques. We present an attractive non-blocking, user-transparent, low overhead checkpointing mechanism for roll-back recovery of failed userapplication processes. 1. Introduction The increasing complexity of real-life problems that are solved using computer technology means that computationally -intensive, time-consuming programs are an increasingly important class of applications. For instance applications that perform neural-network training, simulated annealing or programs that calculate long-term climate forecast might take anything from few hours to several days to execute [1]. Although computer systems are ge...