The Cost of Conservative Synchronization in Parallel Discrete Event Simulations |
Common terms and phrases
activities caused activities occurring analytic approximated lower bound assume Asys average number average per-event overhead begin event caused activities completion messages computed conservative synchronization constant factor define 8;(t described DISCRETE EVENT SIMULATIONS Dmin E A(wn enqueued event creation rate event list algorithm events per window events processed events with time-stamps factor of optimal further messages gate delay hazard rate function hypercube inequality Intel iPSC/2 distributed iPSC/2 distributed memory Langley Research Center lookahead bound lookahead calculation LP's mean window width minimum completion model elements multiprocessor non-queueing number of activities number of events number of objects O₁ optimized serial parallel discrete-event simulation performance analysis Petri net pre-sent priority queue problems processed each window processed per window processor idle random set S₁ serial implementation serial simulation server simulation activity simulation model Stochastic Dominance stochastic portion subset synchronization protocol Theorem 4.2 time-stamps less topology Us(j w₁ Warp Xj,k