S. Skold and R. Ronngren. Event sensitive state saving in Time Warp parallel discrete event simulation. In Proc. of the Winter Simulation Conference, pages 653--660. Society for Computer Simulation, Dec. 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....a periodic basis. In addition, most of these models [12, 14, 20] assume there exists a fixed value for the time to record a state as a checkpoint (which is usually a good approximation) and that the granularity of simulation events has small variance. A more general model is the one presented in [22], which takes into account how the exact granularity of simulation events affects the coasting forward time, and thus the state recovery time. The relevance of this model is in that several real world simulations, such as battlefield simulations or simulations of personal communication systems, ....

....to be restored due to a future rollback. P (S) will be used in the construction of the cost model expressing the checkpointing recovery overhead associated with S. Denoting with ffi s the time to save or reload the LP state vector, which is assumed to be constant as in most previous analyses (see [12, 14, 20, 22, 23]) the checkpointing overhead C(S) associated with the state S can be expressed as follows ( 3 ) C(S) ffi s if S is recorded 0 if S is not recorded (1) Expression (1) points out that if S is recorded as a checkpoint then there is a checkpointing overhead associated with it which is ....

[Article contains additional citation context not shown here]

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations ", Proc.

....section 5 contains our concluding remarks and some discussions on future work. 2 Related Work Over the years, many schemes such as incremental, infrequent, adaptive or hybrid have been proposed to reduce the state saving cost. Instead of saving the entire vector, the incremental approach [3, 18, 22] saves only those changes to the state. When memory consumption is concerned, the incremental approach is useful if the size of state vector is large and only a small portion is modified after an event has been executed. However, the incremental approach requires additional processor time to ....

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulation", Proc. of 1996 Winter Simulation Conference, December 1996.

....from a previous saved one and replying intermediate tasks in a coasting forward, which adds to recovery a time penalty. Several policies for selecting the states to be saved have been proposed in order to achieve good tradeoffs between the time spent for checkpointing and the recovery latency [2, 14, 15, 18, 22]. Some of them cope also with variable rollback rate in the lifetime of the computation [7, 16, 19, 20] A rather different scheme relies on a checkpointing protocol known as incremental state saving [1, 23, 25] Compared to copy state saving, this protocol reduces both the checkpointing overhead ....

....benchmark for two main reasons: i) its parameters (e.g. task processing time, dimension of the state, etc. can be easily modified, ii) in the context of optimistic parallel discrete event simulation it is one of the most used benchmarks for testing performance of checkpointing recovery schemes [4, 18, 20, 22, 26]. Furthermore, the PHOLD model shows a rollback behavior similar to many other synthetic benchmarks and also to several real world models, therefore a wide class of simulation models would lead to performance data quite similar to those obtained with it. We have considered two different pairs of ....

[Article contains additional citation context not shown here]

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations ", Proc. 1996 Winter Simulation Conference, December 1996.

....a periodic basis) In addition, most of these models [11, 13, 19] assume there exists a fixed value for the time to record a state as a checkpoint (which is usually a good approximation) and that the granularity of simulation events has small variance. A more general model is the one presented in [21], which takes into account how the exact granularity of simulation events affects the coasting forward time (thus the state recovery time) The relevance of this model relies on the fact that several real world simulations, such as battlefield simulations or simulations of mobile communication ....

....rollback. The value of P (S) will be used in the construction of the cost model which expresses the checkpointing recovery overhead associated to S. Denoting with ffi s the time to save or reload a whole LP state, which is therefore assumed to be constant as in most previous analyses (see [11, 13, 19, 21, 22]) the checkpointing overhead C(S) associated to the state S can be expressed as follows ( 3 ) C(S) 8 : ffi s if S is recorded 0 if S is not recorded (1) Expression (1) points out that if S is recorded as a checkpoint then there is a checkpointing overhead associated to it which is ....

[Article contains additional citation context not shown here]

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations ", Proc. of 1996 Winter Simulation Conference, December 1996.

....updated at each event execution and the rollback length are minimal. ing to a rollback cost proportional to the average distance between checkpoints. In these approaches the tradeoff between the checkpointing frequency and the checkpointing rollback overhead has been the object of several studies [1, 13, 15, 16, 19, 20]. Furthermore, in the case of periodic state saving, adaptive techniques have been introduced to dynamically recalculate the value of the checkpoint interval. In [14, 17, 19] the recalculation is based on the variation of the rollback behavior (i.e. rollback frequency and rollback length) of the ....

....no productive simulation work to be performed. In the case of periodic state saving, several analytical models have been proposed to describe the simulation execution time of an LP as a function of its checkpoint interval [13, 15, 19] Among those models, particularly interesting is the one in [20], which takes into account how the granularities of different event types affect the average regeneration time of a missing state. The extended experimental study in [16] pointed out that, under atomic coasting forward, the variation of the checkpoint interval strongly affects the rollback ....

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations ", Proc. 1996 Winter Simulation Conference, December 1996.

....that state end the re execution of intermediate events (coasting forward) in order to re update state variables. Several policies for selecting the states to be saved have been proposed in order to achieve good tradeoffs between the time spent saving state information and the coasting forward time [2, 11, 13, 16, 19]. Some of them [4, 12, 17] also cope with dynamic simulation conditions and with thrashing and throttling phenomena [14] Another software alternative is incremental state saving [1, 20, 21] It maintains a history of before images of the state variables that are modified during event execution. ....

.... Hence, if the average value of the running time of the checkpointing protocol is required in order to use a given selection policy (this always happens in the case of policies based on analytical models describing the execution time of an LP as a function of the checkpointing rollback overhead [11, 13, 17, 19]) then it must be estimated on line through appropriate monitoring. 4 Performance Measures In this section we report some experimental results demonstrating the effectiveness of FSC. We consider five different checkpointing schemes: 1) copy state saving (CSS) 2) optimal sparse state saving ....

[Article contains additional citation context not shown here]

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations", Proc. 1996 Winter Simulation Conference, December 1996.

....For the case of sparse state saving based on simulation events (i.e. each event executions) several analytical models have been proposed to determine the time optimal value of the checkpoint interval. The models in [11, 13, 17] assume little variance of the event execution time. The model in [18] copes with simulations where previous assumption is not verified, hence generating a more general solution. Adaptive schemes for selecting the checkpoint interval have also been presented [4, 12, 17] in order to tackle both thrashing and throttling phenomena [14] and possible dynamic simulation ....

S. Skold and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations", Proc. of 1996 Winter Simulation Conference, December 1996.

No context found.

S. Skold and R. Ronngren. Event sensitive state saving in Time Warp parallel discrete event simulation. In Proc. of the Winter Simulation Conference, pages 653--660. Society for Computer Simulation, Dec. 1996.

No context found.

S. Skold and R. Ronngren. Event sensitive state saving in Time Warp parallel discrete event simulation. In Proc. of the Winter Simulation Conference, pages 653--660. Society for Computer Simulation, Dec. 1996.

No context found.

S. Skold, and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations", Proc. 1996.

No context found.

S. Skold, and R. Ronngren, "Event Sensitive State Saving in Time Warp Parallel Discrete Event Simulations ", Proc. of 1996.

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