TY - GEN
T1 - Extracting more intra-transaction parallelism with work stealing for OLTP workloads
AU - Zhou, Xiaozhou
AU - Wang, Zhaoguo
AU - Chen, Rong
AU - Chen, Haibo
AU - Li, Jinyang
N1 - Publisher Copyright:
© 2017 ACM.
PY - 2017/9/2
Y1 - 2017/9/2
N2 - Online transaction processing systems use two-phase locking (2PL) to guarantee serializability. However, traditional 2PL does not perform well under high contention, because a transaction will be blocked when it fails to acquire lock. This paper proposes a scalable work stealing algorithm for 2PL to leverage intra-transaction parallelism. The key idea is to parallelize the lock holder's work among lock waiters. Compared to traditional 2PL, our approach can achieve up to 2.8X throughput improvement for TPC-C new-order transactions under high contention.
AB - Online transaction processing systems use two-phase locking (2PL) to guarantee serializability. However, traditional 2PL does not perform well under high contention, because a transaction will be blocked when it fails to acquire lock. This paper proposes a scalable work stealing algorithm for 2PL to leverage intra-transaction parallelism. The key idea is to parallelize the lock holder's work among lock waiters. Compared to traditional 2PL, our approach can achieve up to 2.8X throughput improvement for TPC-C new-order transactions under high contention.
UR - http://www.scopus.com/inward/record.url?scp=85030554266&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85030554266&partnerID=8YFLogxK
U2 - 10.1145/3124680.3124748
DO - 10.1145/3124680.3124748
M3 - Conference contribution
AN - SCOPUS:85030554266
T3 - Proceedings of the 8th Asia-Pacific Workshop on Systems, APSys 2017
BT - Proceedings of the 8th Asia-Pacific Workshop on Systems, APSys 2017
PB - Association for Computing Machinery, Inc
T2 - 8th ACM Asia Pacific Conference on Systems, APSys 2017
Y2 - 2 September 2017
ER -