Wolf: a rollback algorithm for optimistic distributed simulation systems
WSC '88 Proceedings of the 20th conference on Winter simulation
The dependence list in time warp
Proceedings of the fifteenth workshop on Parallel and distributed simulation
Causality representation and cancellation mechanism in time warp simulations
Proceedings of the fifteenth workshop on Parallel and distributed simulation
Parallel and Distribution Simulation Systems
Parallel and Distribution Simulation Systems
Early cancellation: an active NIC optimization for time-warp
Proceedings of the sixteenth workshop on Parallel and distributed simulation
Batch based cancellation: a rollback optimal cancellation scheme in time warp simulations
Proceedings of the eighteenth workshop on Parallel and distributed simulation
| Hi-index | 0.00 |
Time Warp parallel simulations need efficient schemes to cancel erroneous event messages in order to achieve acceptable performance levels. Lazy cancellation depends on delaying the cancellation of incorrect computations until rolled-back events are re-processed. An antimessage is then sent for cancellation only if an output message reproduced after rollback is found to be unequal to its counterpart produced before rollback. In this paper, we suggest an enhanced cancellation technique that aims to throttle incorrect computations by asking logical processes (LPs) that have received output messages generated from the execution of an input event that has later been cancelled to mark those messages as temporarily blocked until it is later decided whether or not they should also be cancelled. The receiving LP is blocked once a marked message is due for processing. After re-executing the input event, the rolled-back LP sends the receiving LP either an "unblock" message or an antimessage for each previously marked output message depending on the result of the equality test of that output message versions before and after rollback. Experimental results for a synthetic benchmark show that our technique can yield better performance than both lazy and aggressive cancellation.