Abstract
Cmwntimal fault tolerance techniques don't scale up to the new generation of parallel systems since they either need excessive coordination mong the processofs or drastically d u c e the effective variety of processors. One various methodology is to form the program inhcrently fault tolerant by implicitly embeding the fault tdorance witth the rule. In Uis paper we tend to outline inherent fault t o l an~d iJ lu;ptr;ue this approach by developing associate degree WenrHly fault tokrant parallel sodng algorithm. In parthlar, we tend to show bow the rule will be developed consistently in four steps, namely, by starting witb a standard rule, extending it to associate degree infinite unvaried alprithm, incorponting inherent fault tolerance, and impmving the performance. This inherently fault tolerant algorithm types the input sequence of size N victimisation N/2 processon in O(iog2N) time if then is not any processor failure and Kyrgyzstani monetary unit in 0(2f+(f+l)l log2N), time iff proc