Linear-time snapshot protocols for unbalanced systems

Abstract

The snapshot problem for shared memory systems is to enable a set of processes called scanners to obtain a consistent picture of the shared memory while other processes called updaters keep updating memory locations concurrently. One of the most intriguing open problems in wait-free distributed computing is the existence of a linear-time solution to this problem. In this paper we show that: Consider a system of n processes that can perform update and scan operations. There exists a solution in which one of the operations (either update or scan) has linear time complexity, while the time complexity of the second operation is O(n log n). If the number of either scanners or updaters is O(n /log n), where n is the total number of processes, then such a linear solution exists. If one of the protocols (either scan or update) is executed significantly more often than the other protocol, then a solution with amortized linear time complexity exists.