unni@sunset.sm.unisys.com (Unni Warrier) (08/01/87)
(I AM ORNA BERRY, orna%israern.bitnet@wiscvm.wisc.edu) (till 8/7/87 at cam.unisys.com) >From jfjr@mitre-bedford.ARPA Fri Jul 31 05:03:23 1987 > simulation paradigms. The major problem with simulations in general > is validation and verification (debugging). Its.... > to junk the roll back method for these reasons. > Jerry Freedman, Jr "As you wander through life I have been involved in simulations and implementations of the Time Warp Mechanism. It is very easy to keep traces of messages (positive and negative, or "messages" and "antimessages") vs. traces of only these messages that were used in the final computation (positive messages that were not cancelled). If the software engineering of the system is decent, one can also get snapshots of every object, and the event queues on all machines, so it is not so hard to develop good debugging tools. I have seen (in the Univ. of Calgary) a graphics interface showing important attributes of the objects in the simulation and of the data structures reflecting the simulation time, and that made the debugging really pleasant. So, all in all, when you understand the mechanism you can construct a debugger that eases the users' life and make it very pleasant to develop code for the TW mechanism. I must point out that we had no semantic limitations on the kind of applications we devised for TW (like systems of M/M/2 queues) which we had difficulties with in other distributed simulation mechanisms. Orna