Surveillance for controlled performance degradation during failures

Graceful performance degradation due to failures can be achieved by maintaining updated system reachability information. This can be accomplished through the early detection of site and communication failures and repairs. Surveillance protocols are responsible for gathering and updating this information which are known as views. The author present a simple, reliable surveillance protocol along with its application to distributed transaction processing systems. The authors examine the effect of surveillance on the performance of transaction processing during failures with and without surveillance. They show that while the cost of surveillance overhead is affordable through a fixed additional MIPS, the benefits of surveillance are indispensable.<>