Lazy fault tolerance-a method for dependable distributed systems
M. Saito, Takanori Yokoyama, Mitsumasa Shimada · 2002
We present a method called Lazy Fault Tolerance for refining the reliability of distributed systems. Lazy Fault Tolerance uses data redundancy and the data of objects are distributed over computers in accordance with their 'nativity', The data of system management objects, which control the whole system based on the information of each computer, are naturally distributed over all computers. The object will restore its data by gathering all information from all computers in the case of failures. Lazy Fault Tolerance can greatly reduce the performance degradation in the absence of failures, although it takes a longer time to restore the system when a computer fails.