Hamnes, D. O. and Tripathi, A. 1994. Evaluation of a local adaptive protocol for distributed simulation. In Proc. 1994 International Conference on Parallel Processing. CRC, 127--134.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to complete the simulation in a time less than the critical path time. Using a previously derived su#cient condition for a PDES algorithm to be super critical, we show that the changepoint PDES algorithm is super critical. Finally, we qualitatively compare previous state based PDES algorithms [8] [9] to the changepoint A less explored viewpoint in PDES, and discrete event simulation in general, is the relationship between simulated events and the set of measures defined on a model. As models become larger and more complex, the subtle result is that there are more events that do not contribute ....

....do not become overly aggressive. 2.5.6 State based Recently, three algorithms have been proposed that can be categorized as state based. They attempt to continuously adjust their optimism based on local state information directly a#ected by the actions of other LPs. The first two algorithms [8] [9] are similar in that they both utilize information about the input channels of an LP to decide when an LP should process events. Hamnes and Tripathi [9] use a control theory feedback approach to determine windows in which an LP should process (receive) events and send messages to other LPs. ....

[Article contains additional citation context not shown here]

D. Hamnes and A. Tripathi, "Evaluation of a local adaptive protocol for distributed discrete event simulation," in Proceedings of the

....first, they are adaptive in that the LP s continually adjust their optimism, and second, adaptive decisions are based on state information which, although available locally, is directly affected by the actions of other LP s. These protocols are similar to NPSI protocols. The first two protocols [HaTr94, FeTr94] are similar to each other in that they both utilize information about channels to decide when and for how long LP s should wait. Intuitively, it may be argued that our approach has an advantage over these because LP s in NPSI protocols can receive infor 22 mation from all of their predecessors ....

....only about immediate predecessors. The benefit comes from the fact that in a channel based protocol, information has to percolate through the predecessors of an LP before it reaches that LP. During this time, the LP may have moved farther ahead than it should have. Moreover, the protocols in [HaTr94, FeTr94] cannot determine if a predecessor is rolling back until they receive a null or anti message conveying that information. NPSI protocols are able to determine this information earlier. The third protocol in this class [DaFu94] uses memory consumption as the basis for limiting optimism. The ....

[Article contains additional citation context not shown here]

Hamnes, D.O. and Tripathi, A., "Evaluation of a local adaptive protocol for distributed discrete event simulation", Proceedings of the 1994 International Conference on Parallel Processing, August 1994, Vol. III, 127-134.

....are truly adaptive in that the LP s continually adjust their optimism, and second, adaptive decisions are based on state information which, although available locally, is directly affected by the actions of other LP s. Thus, these protocols are similar to NPSI protocols. The first two of these [HaTr94, FeTr94] are similar to each other in that they both utilize channel information to decide when and for how long LP s should wait. Intuitively, it may be argued that our approach has an advantage over these because LP s in NPSI protocols can receive information from all of their predecessors whereas ....

....information only about immediate predecessors. The benefit comes from the fact that in a channel protocol, information has to percolate through the predecessors of an LP before it reaches that LP. During this time, the LP may have moved farther ahead than it should have. Moreover, the protocols in [HaTr94, FeTr94] cannot determine if a predecessor is rolling back until they receive a null or anti message conveying that information. NPSI protocols are able to determine this information earlier. The third protocol in this class [DaFu94] uses memory consumption as the basis for limiting optimism. The ....

[Article contains additional citation context not shown here]

Hamnes, D.O. and Tripathi, A., "Evaluation of a local adaptive protocol for distributed discrete event simulation", Proceedings of the 1994 International Conference on Parallel Processing, August 1994, Vol. III, 127-134.

....of other LP s. Similarly, in Adaptive Time Warp [BaHo90] an LP may decide to block after executing an event based on local history and statistical estimation. In [Madi93] LP s estimate each others logical clock values and block if their clock value differs largely from that of another LP. In [HaTr94], a real time blocking window is computed each time an LP executes an event and the LP blocks for an amount of time equal to this window (which may be zero) Similarly, in [FeTr94] an LP blocks probabilistically for some amount of time after each event execution. The new class of adaptive ....

Hamnes, D.O. and Tripathi, A., "Evaluation of a local adaptive protocol for distributed discrete event simulation", Proceedings of the 1994 International Conference on Parallel Processing, August 1994, Vol. III, 127-134.

....of optimistically processing an event an LP is blocked. The blocking interval is based either on statistical estimates of the next event arrival [4, 1] computing tradeoffs between rollback and blocking [5] global system state (NPSI [23] or probabilistically [6] The Local Adaptive Protocol [10] computes real time blocking windows based on previous LVT increments, but is prone to deadlock. Another class of adaptive schemes use memory management for optimism control. Das s adaptive memory scheme [3] AMM) based on Cancelback [12] rolls back LPs to reclaim event memory. The scheme ....

D. O. Hamnes and A. Tripathi. Evaluation of a local adaptive protocol for distributed discrete event simulation. In Proceedings of the 1994 International Conference on Parallel Processing, pages III:127--134, August 1994.

No context found.

Hamnes, D. O. and Tripathi, A. 1994. Evaluation of a local adaptive protocol for distributed simulation. In Proc. 1994 International Conference on Parallel Processing. CRC, 127--134.

No context found.

D. O. Hamnes and A. Tripathi. Evaluation of a local adaptive protocol for distributed discrete event simulation. In Proceedings of the 1994 International Conference on Parallel Processing, pages III:127--134, August 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC