D.M. Nicol, "The Cost of Conservative Synchronization in Parallel Discrete Event Simulations", Journal of the Association for Computing Machinery, pp. 304-333, 1993.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....is usually a barrier synchronization such that all LPs complete the current computation phase before proceeding to the next one. Thus, these algorithms are accurate, nonaggressive, riskfree and synchronous. Examples include lookahead based windows [18] 19] Conditional Events [20] and YAWNS [21]. The main limitation of synchronous algorithms is the determination of the size of the logical time window. The optimal size depends on the application and user expertise. Performance can be bad if the window is either too small or large. Furthermore, these algorithms su#er some of the same ....

D. Nicol, "The cost of conservative synchronization in parallel discrete event simulations, " Journal of the ACM, vol. 40, pp. 304--333, April 1993.

....null messages only upon demand; an LP requests a null message from the LP which sends messages to an empty input queue [25] In the carrier null message approach [27] lookahead information is attached to a null message. Another form of deadlock avoidance is the conservative time window approach [17, 26] in which time windows are successively generated with the property that all of the events within each of the windows are safe to process. The deadlock detection and breaking approach relies upon algorithms to detect deadlocks. In [38] a knot detection algorithm is used to detect the deadlock, ....

D.Nicol. The cost of conservative synchronization in parallel discrete event simulations, Journal of the ACM 40(2), pp. 304-333.

....algorithm is synchronous (i.e. all LPs carry out local computations followed by a global computation) and the performance studies carried out in the paper assume a network with very high number of jobs. The paper does not compare the performance of the conditional event protocol with others. Nicol[15] describes the overheads of a synchronous algorithm similar to the conditional event algo24 rithm studied in this paper. A mathematical model is constructed to qualify the overhead due to various protocol specific factors. However, for analytical tractability, the paper ignores some costs for ....

D.M. Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the ACM, 40(2):304--333, April 1993.

....CMB is an asynchronous algorithm; null messages function as time advancement noti cations between logical processes. Several synchronous algorithms exist, including Chandy and Sherman s conditional event approach [23] Lubachevsky s bounded lag algorithm [71, 72, 73] and Nicol s YAWNS protocol [86, 87]. Synchronous 9 approaches exploit the computational eciency of global parallel operations like barriers and min reductions. They are valued for their simplicity as well as provable scalability. The rst optimistic parallel simulation protocol is Je erson s Time Warp algorithm [51] It was ....

.... Synchronization The initial implementation of DaSSF follows and extends the synchronization scheme of a previous parallel simulator called Nops [103] which is a window based barrier synchronization scheme extended from Nicol s YAWNS protocol [86] The protocol has been shown to be scalable [87, 90]. Each timeline is implemented as a logical process with its own event list. In addition, each processor maintains a future event list as well. At the point an event is written to a channel that mapped between two distinct timelines, the event is added to the future event list of the current ....

[Article contains additional citation context not shown here]

D. M. Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the ACM, 40(2):304-33, April 1993.

....advances the local simulation time beyond the computed LBTS value, and if no simulator sends an event message in the simulated past, then it can be guaranteed that no simulator will receive events in their local view of the simulated past. A number of approaches to computing the LBTS exist [14 18]. The backplane makes use of time management services provided by the RTIKIT [19] which uses a butterfly barrier technique first proposed by Brooks [20] Using the butterfly barrier, simulators exchange local LBTS and message count information with each other in a series of rounds. After the ....

Nicol, D.M., The Cost of Conservative Synchronization in Parallel Discrete Event Simulations. Journal of the Association for Computing Machinery, 1993. 40(2): p. 304-333.

....a barrier synchronization such that all LP s complete the current computation phase before proceeding to the next one. Thus, these protocols are accurate, non aggressive, without risk and synchronous. Examples include lookahead based windows [Luba89, AyRa92] Conditional Events [ChSh89] and YAWNS [Nico93]. 16 The main limitation of synchronous protocols is the determination of the size of the logical time window. The optimal size depends on the application and its identification requires user expertise. Performance can be very bad when the window is either too small or too large. Further, these ....

Nicol, D.M., "The cost of conservative synchronization in parallel discrete event simulations", Journal of the ACM, Vol. 40, No. 2, April 1993, 304-333.

....of the two variations, a relative short presimulation is performed to identify those times, and only those times, at which the simulation algorithm requires processors pairs to synchronize. Finally, the cost of conservative synchronization in parallel discrete event simulations is discussed in [14]. This paper analytically studies the performance of a synchronous conservative parallel discrete event simulation protocol. The class of models considered simulates activity in a physical domain, and possesses a limited ability to predict future behaviour. Using a stochastic model, it is shown ....

D. M. Nicol. The Cost of Conservative Synchronization in Parallel Discrete Event Simulations. Journal of the Association for Computing Machinery, 40(2):304--333, 1993.

....a common framework. The work discussed here concerns the empirical comparative analysis of protocols when applied to a set of applications in a common environment. To date, other simulation studies have been fairly narrowly focused and inconclusive. Analytic results are beginning to appear (e.g. [Nico89]) but they do not consider important application characteristics. Our hypothesis, shared by many, is that the effectiveness of parallel simulation protocols will be highly dependent on the applications using them. Thus, an open question concerns the determination of the suitability of classes of ....

Nicol, D.M., "The Cost of Conservative Synchronization in Parallel Discrete Event Simulations", unpublished manuscript, June, 1989.

....Also, changing the existing functionality or configuration is generally time consuming. Furthermore, it is hard to use these testbeds to determine whether their control systems scale well to networks of larger size. An alternative approach to study network design problems is using simulators [1,3,9,10,11]. The advantage of simulation environments over testbeds is that they allow for easy monitoring of the system state and rapid modification of the functionality of the simulated system. While these simulators are suited for studying the behavior and performance of a specific controller or an ....

....or equal to T G . Otherwise, the event is buffered in the event queue. T G is computed as the minimum of T S and T W , which are explained below. T S is the simulation time as computed by the causalitycontrol protocol, which is based on the window based synchronization protocol described in [11]. T W , the scaled wall clock time, is a linear function of the processor time taken from a reference processor, i.e. T W = Ptime K ratio , where P time denotes the processor time and K ratio is a control parameter that can be changed by the user. Performing the simulation in the above ....

[Article contains additional citation context not shown here]

D.M. Nicol, "The cost of conservative synchronization in parallel discrete-event simulations," Journal of ACM, 40(2):304-333, April 1993

....(LBTS) within a parallel discrete event simulation. For this discussion, we define the LBTS to be the minimum timestamp on any message that can possibly be received in the future. The LBTS algorithm is a fundamental computation used by conservative synchronization protocols, such as YAWNS[5] and the conditional event protocol[2] as well as optimistic protocols when computing a Global Virtual Time (GVT) Our two algorithms assume differing network service models, and we show that the service model assumptions and choices can have a large impact on overall performance. The main ....

D. M Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the Association for Computing Machinery, 40(2):304--333, 1993.

....time delay between any two simulation objects. In the SPEEDES simulation engine, Steinman[6] utilizes the Time Buckets Protocol in which processes peri odically resynchronize to determine a lower bound on events which are safe to process. Time Buckets is also similar to our approach. Nicol[7] describes the YAWNS protocol which is also similar to our approach. The main contributions of our approach are threefold. First we utilize the Broadcast capability of an Ethernet network to give an efficient, albeit unreliable, one to many message dissemination capability. Secondly, we use ....

D. M. Nicol, "The cost of conservative synchronization in parallel discrete event simulations," Journal of the Association for Computing Machinery, vol. 40, no. 2, pp. 304--333, 1993.

....and the details are covered in Section 2.1 and Figure 2. LP i .MinTime is in turn used to compute GST (part 2(G) the smallest timestamp of all events in the system. GST is also used in the safetime computation. 2. 1 Safetime Computation There are different ways of computing the safetime values [1, 2, 5]. We initially used the algorithm described in [1] and later came up with other methods to compute the safetimes [4] The latter methods use fewer number of supersteps. The safetime computation method described in this paper further reduces the number of super steps 2 . We note that LP i s ....

D. M. Nicol. The cost of conservative synchronization in parallel discrete-event simulation. Journal of the ACM, 40(2):304--333, April 1993.

....Coherency protocols complicate the simulation problem considerably, but are a facet we need not deal with currently. The second difference is the synchronization mechanism used to maintain the fidelity of the parallel simulation. WWT uses a special case of the YAWNS synchronization protocol [7] while the original version of LAPSE uses a synchronization mechanism combining YAWNS with appointments [8] As noted, we are modifying the synchronization protocol of LAPSE to estimate performance for codes executing on workstation clusters. We will present a more thorough literature review in ....

Nicol, D. The cost of conservative synchronization in parallel discrete-event simulations. Journal of the ACM, 40(2):304--333, April 1993.

....they must be scheduled. We do expect there to be some benefit from this, since there are relatively few inter cluster channels. The probability of an event arriving via an inter cluster channel at any specific simulation time is relatively low. 5.1. 1 Conservative techniques The YAWNS protocol [48] suggests another statically scheduled algorithm which is shown in [56] to perform well with large network simulations. While our mechanism requires knowledge of a minimum external channel delay for each cluster, the YAWNS protocol requires only the minimum delay, #, over all inter cluster ....

David M. Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the Association for Computing Machinery, 40(2):304--333, April 1993.

No context found.

D.M. Nicol. The cost of conservative synchronization in parallel discrete-event simulations. Journal of the ACM, 40(2):304--333, April 1993.

....communication networks. The present work is part of our on going efforts to optimize DaSSF s performance, while maintaining or increasing its flexibility. The literature on conservative synchronization of parallel discrete event simulations describes synchronous approaches based on barriers (e.g. [1, 12, 13]) and asynchronous approaches which govern a submodel s advance as a function of the advance of other submodels which may affect it (e.g. 2] Synchronous approaches exploit the computational efficiency of global parallel operations like barriers and min reductions, and are valued for their ....

D. Nicol. The cost of conservative synchronization in parallel discrete-event simulations. Journal of the ACM, 40(2):304--333, April 1993.

....code is altered to cause WWT application processes to synchronize every B cycles. Any communication is deferred until the next barrier with the assurance that the barrier occurs before the communication can have affected its recipient. This method of synchronization is a special case of the YAWNS [12, 14] protocol. The WWT ignores any network contention by assuming that the latency of every message is fixed, and known. By contrast, Paragon processors are less tightly coupled. In a cache coherent setting any memory reference might generate a network event; in a message passing setting only an ....

....who recognizes the time stepping criteria enters zero into the lookahead bound reduction. Seeing zero emerge from the reduction, processors know to time step. 3.5.4 Synchronization for the simple network Synchronization using the simple network model is also simple. We have implemented the YAWNS [12, 14] window based protocol. Assume that all processors are synchronized at time t. For each VP we find a lower bound on the next time a traveler from that VP enters the network (the application lookahead) and add to that a message latency. The latency is the time it takes the header of the message ....

D.M. Nicol. The cost of conservative synchronization in parallel discrete-event simulations. Journal of the ACM, 40(2):304--333, April 1993.

No context found.

D.M. Nicol, "The Cost of Conservative Synchronization in Parallel Discrete Event Simulations", Journal of the Association for Computing Machinery, pp. 304-333, 1993.

No context found.

D.M. Nicol. "The cost of conservative synchronization in parallel discrete event simulations". Journal of the ACM, 40(2):304--333, April 1993.

No context found.

D.M. Nicol. "The cost of conservative synchronization in parallel discrete event simulations". Journal of the ACM, 40(2):304--333, April 1993.

No context found.

D.M. Nicol. \The cost of conservative synchronization in parallel discrete event simulations". Journal of the ACM, 40(2):304-333, April 1993.

No context found.

D.M. Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the ACM, 40(2):304--333, April 1993.

No context found.

D.M. Nicol. "The cost of conservative synchronization in parallel discrete event simulations". Journal of the ACM, 40(2):304--333, April 1993.

No context found.

D.M. Nicol. "The cost of conservative synchronization in parallel discrete event simulations ". Journal of the ACM, 40(2):304--333, April 1993.

No context found.

David M. Nicol. The cost of conservative synchronization in parallel discrete event simulations. Journal of the ACM, 40(2):304--333, April 1993.

First 50 documents

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