Time and message bounds for election in synchronous and asynchronous complete networks
Time and message bounds for election in synchronous and asynchronous complete networks
复制标题
同步和异步完整网络中选举的时间和消息范围
DOI:
10.1145/323596.323613
复制
发表时间:
1985
期刊:
影响因子:
--
通讯作者:
E. Gafni
中科院分区:
文献类型:
--
作者:
Y. Afek;E. Gafni
This paper addresses the problem of distributively electing a leader in both syn- chronous and asynchronous complete networks. O(n log n) messages synchronous and asynchronous algorithms are presented. The time complexity of the synchronous algorithm is O(log n), while that of the asynchronous algorithm is O(n). In the synchronous case, a lower bound of 12(nlogn) on the message complexity is proven. It is also proven that any message-optimal synchronous algo- rithm requires 12(log n) time. In proving these bounds, the type of operations performed by nodes are not restricted. The bounds thus apply to general algorithms and not just to comparison-based algorithms. to be selected from a set of nodes which initially differ only in their identifiers (ids), with no node being aware of any other id. An arbitrary subset of nodes wakes up spontaneously at arbitrary times and starts the algorithm by sending messages over the network. When the message exchange terminates, a leader is distinguished from all other nodes. This paper specializes in the problem of electing a leader in a complete network. In such a network, each pair of nodes is connected by a bidirectional communication link. Before the algorithm starts, no node has any information on any of the other nodes. Hence, the incident links of a node, on which no message was sent or received, are indistinguishable. We consider both synchronous and asynchronous modes of