On Closed Nesting and Checkpointing in Fault-Tolerant Distributed Transactional Memory
Aditya Dhoke, Binoy Ravindran, Bo Zhang · 2013
We consider the closed nesting and checkpointing model for transactions in fault-tolerant distributed transactional memory (DTM). The closed nested model allows inner-nested transactions to be aborted (in the event of a transactional conflict) without aborting the parent transaction, while checkpointing allows transactions to rollback to a previous execution state, potentially improving concurrency over flat nesting. We consider a quorum-based replicated model for fault-tolerant DTM, and present algorithms to support closed nesting and checkpointing. The algorithms use incremental validation to avoid communication overhead on commit, and ensure 1-copy equivalence. Our experimental studies using a Java DTM implementation of the algorithms on micro and macro benchmarks reveal the conditions when they improve transactional throughput over flat nesting, and also their relative advantages and disadvantages.