Dynamic Cancellation: Selecting Time Warp Cancellation Strategies at Runtime

Abstract

The performance of Time Warp parallel discrete event simulators can be affected by the cancellation strategy used to send anti-messages. Under aggressive cancellation, antimessage generation occurs immediately after a straggler message is detected. In contrast, lazy cancellation delays the sending of anti-messages until forward processing from a straggler message confirms that the premature computation did indeed generate an incorrect message. Previous studies have shown that neither approach is clearly superior to the other in all cases (even within the same application domain). Furthermore, no strategy exists to make a priori determination of the more favorable cancellation strategy. Most existing Time Warp systems merely provide a switch for the user to select the cancellation strategy employed. This paper explores the use of simulation time decision procedures to select cancellation strategies. The approach is termed Dynamic Cancellation and it assigns the capability for selecting cancellation strategies to the Logical Processes (LPs) in a Time Warp simulation. Thus, within a single parallel simulation both strategies may be employed by distinct LPs and even across the simulation lifetime of an LP. Empirical analysis using several control strategies show that dynamic cancellation always performs with the best static strategy and, in some cases, dynamic cancellation provides some nominal (5–10%) performance gain over the best static strategy.