An overhead reducing technique for Time Warp

Using consistent global checkpoints to synchronize processes in distributed simulation

“…The line 3 of matrix M is replaced by the dependencies vector received and a new recovery line, represented by vetor [1, 1, 2, 1], is extracted from matrix. This procedure is repeated for checkpoints C 2 2 and C 4 2 , but when receiving information of occurrence of the checkpoint C 3 3 , by means of the forth message, the Matrix M becomes: Only when receiving the sixth message, the observer process will be able to identify the recovery line [2,2,4,2].…”

“…A similar situation occurs in the process P 4 , which does not need to carry out rollback. When the process P 2 receives all replies, it is able to identify the recovery line in which the computation was reconstituted, that is, [2,2,4,3]. Finally, this line is sent to all other processes, by means the message FinishRollback, and the epoch is advanced.…”

“…Many papers have been published trying to find new solutions for some traditional problems such as new mechanisms for memory management [3,12] and modifications in the rollback procedure [20]. Changes on the protocol were also proposed such as the introduction of a conexionoriented communication-mechanism among the logical processes [10], or the possibility of saving states concurrently to the logical processes activities [18].…”

Rollback Solidário: um novo protocolo otimista para simulação distribuída

“…Desde então, diversos trabalhos foram desenvolvidos com o objetivo de melhorar o desempenho desse protocolo. Com o passar dos anos vários problemas foram identificados e soluções foram apresentadas.Ainda hoje, vários artigos têm sido publicados tentando encontrar novas soluções para alguns dos seus tradicionais problemas como, por exemplo, novos mecanismos para o gerenciamento da memória(VEE;HSU, 2002;CHUNG;XU, 2002;LI;TROPPER, 2004) e modiiicações no procedimento de rollback (ZHANG; TROPPER, 2001; ZENG; CAI; TURNER,…”