On Reducing False Conflicts in Distributed Transactional Data Structures

We present techniques for reducing false conflicts in distributed transactional data structure (DDS). The open nesting transactional model is the common solution because it allows nested transactions to commit independently of their parent transaction, thereby objects in the transaction read-set and write-set are released early, minimizing aborts due to false conflicts and improving concurrency. We present three protocols for avoiding false conflicts in DDS. Our first protocol, QR-ON, incorporates open nesting into the QR protocol that manages concurrency control for distributed transactional memory systems using quorum-based replication. We then introduce Optimistic Open Nesting, QR-OON, in which open-nested transactions commit asynchronously so that subsequent transactions can proceed without waiting for the commit of previous transactions. Finally, we propose an early release methodology, QR-ER, in which objects that do not affect the final state of the shared data are dropped from transaction's read-set, which avoids false conflicts and reduces communication costs. Our implementation and experimental studies revealed that QR-OON outperforms QR-ON by up to 43%, and that QR-ER outperforms QR-ON and QR-OON by up to 10X.