An experimental analysis of replicated copy control site failure and recovery

The authors examine the effect of failures on the consistency of replicated copies, measure the rate at which inconsistency can be removed, and measure the overheads associated with replicated copy control. They have extended the prototype system RAID to utilize the ideas of session numbers, nominal session vectors, fail-locks, and control transactions for replicated copy control. The authors discuss the results of three experiments. In the first experiment the authors measured the overhead for fail-locks maintenance, the overhead for copier transactions. In the second experiment they observed the effects of site failure on data availability. In the last experiment they examined the maintenance of consistency of replicated copies during site failure and recovery