Impact of rollback recovery to program locality of reference and page replacement policies
Ying–Dar Lin, S. Bhattacharya · 2002
We study the impact of checkpointing-based fault tolerance on the program locality of reference and memory page (or cache block) replacement policies. We demonstrate that the current research in page/block replacement algorithms may not operate well with fault-tolerant programs. We include the identification of this critical research topic, and the development of an analytical model for fault propagation in a checkpointed program. This model is corroborated using simulation, and we propose a new page/block replacement algorithm that takes into account fault-triggered rollback(s). Performance is shown by an example study.