On rolling back and checkpointing in time warp

Abstract

AbstractÐIn this paper, we present a family of three algorithms which serve to perform checkpoints and to roll back Time Warp. These algorithms are primarily intended for use in simulations in which there are a large number of LPs and in which events have a small computational granularity. Important representatives of this class are VLSI and computer network simulations. In each of our algorithms, LPs are gathered into clusters via algorithms which are application dependent. In order to examine the performance of our algorithms and to compare them to Time Warp, we made use of two of the largest digital logic circuits available from the ISCAS89 benchmark series of combinational circuits. The execution time, number of states saved, and maximal memory consumption were compared to the same quantities for Time Warp. Our results indicated that each of the algorithms occupies a different point in the spectrum of possible trade-offs between memory usage and execution time, ranging from substantial memory savings (at a comparable cost in speed) to memory savings and a comparable speed to Time Warp. Hence, an important benefit of our algorithms is the ability to trade off memory requirements with execution time.

Synchronization methods in parallel and distributed discrete-event simulation

“…Since both mechanisms use infrequent state saving, coast-forward operations are required during rollbacks with the associated overhead. Using several digital circuit models as benchmarks, different combinations of these rollback and checkpointing mechanisms have been evaluated quantitatively [108]. The experiments showed that, while memory usage can be reduced by up to 40% in some cases, the execution time is comparable or even worse than obtained with the original TW protocol, indicating that a trade-off must be made between execution efficiency and memory conservation.…”

XTW: a parallel and distributed logic simulator

“…A cluster level event queue (CLEQ) is part of XEQ and stores events which are sent to other clusters. XTW is an outgrowth of CTW [2]. Three techniques for checkpointing and rolling back in CTW are described in [2], each occupying a different point in a memory vs. execution time trade-off.…”

“…XTW is an outgrowth of CTW [2]. Three techniques for checkpointing and rolling back in CTW are described in [2], each occupying a different point in a memory vs. execution time trade-off. XTW makes use of one of these techniques, local rollback,local checkpoint.…”

“…In this subsection, we present results comparing the performance of XTW and CTW [2] [3]. We compare XTW to CTW because CTW is oriented towards logic simulation, and exhibits a superior performance to Time Warp [2] in this domain.…”

“…Logic and behavioral simulation is a low granularity and tightly coupled computational task which poses a significant challenge to the development of a distributed simulator. Clustered Time Warp (CTW) [2], [3] was developed with these problems in mind. As the name implies, in CTW LPs (representing gates) are gathered into clusters.…”