DTranx: A SEDA-based Distributed and Transactional Key Value Store with Persistent Memory Log
Ning Gao, Zhang Liu, Dirk Grunwald

TL;DR
DTranx is a distributed key-value store leveraging persistent memory and advanced concurrency techniques to achieve high throughput and efficiency without sacrificing consistency.
Contribution
It introduces a persistent memory aware log, a state transition based garbage collection, and a hybrid commit protocol for improved performance and consistency.
Findings
Achieves 514.11k transactions/sec with 36 servers and 95% read workload.
Persistent memory log is 30 times faster than SSD-based systems.
Garbage collection is efficient, non-disruptive, and maintains low log space usage.
Abstract
Current distributed key value stores achieve scalability by trading off consistency. As persistent memory technologies evolve tremendously, it is not necessary to sacrifice consistency for performance. This paper proposes DTranx, a distributed key value store based on a persistent memory aware log. DTranx integrates a state transition based garbage collection mechanism in the log design to effectively and efficiently reclaim old logs. In addition, DTranx adopts the SEDA architecture to exploit higher concurrency in multi-core environments and employs the optimal core binding strategy to minimize context switch overhead. Moreover, we customize a hybrid commit protocol that combines optimistic concurrency control and two-phase commit to reduce critical section of distributed locking and introduce a locking mechanism to avoid deadlocks and livelocks. In our evaluations, DTranx reaches…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsDistributed systems and fault tolerance · Cloud Data Security Solutions · Security and Verification in Computing
