2011 Sixth International Symposium on Parallel Computing in Electrical Engineering (2011)
Luton, United Kingdom
Apr. 3, 2011 to Apr. 7, 2011
. In addition to trade-offs between convergence and response spans as described by Gouda and Evangelist , we consider state space utilization with regard to processing and communications, and sensitivity to delay time estimates. In order to be delay insensitive, estimates of transmission and/or computation time are used, which we implement as virtual nodes. The estimated number of nodes is defined as NE. We define M as the number of processes, and N as the total number of nodes required in the problem space. These algorithms require that NE > N in order that correctness and desired behavior can be guaranteed. We define auto-synchronization as the characteristic of a system such that performance is related to N, rather than NE, even though state space necessarily is related to NE. Although our examples are delay insensitive deterministic unidirectional token rings, we believe that these concepts apply to many other paradigms of self-stabilization.
Algorithms, parallel, distributed computing, mutual exclusion, self-stabilization, synchronization, token rings
F. Haddix, "Trade-offs in Token Ring Systems," 2011 Sixth International Symposium on Parallel Computing in Electrical Engineering(PARELEC), Luton, United Kingdom, 2011, pp. 189-194.