π€ AI Summary
This work addresses performance degradation in high-load key-value storage systems, where resource contention frequently causes thread blocking, context switches, or stalls. To mitigate these issues, the authors propose a novel lock-free concurrency control mechanism based on a βnotificationβ paradigm, integrated with incremental record updates. This approach effectively eliminates thread stalls and context switches along critical paths such as B-tree index maintenance. By substantially reducing futile retries common in conventional lock-free operations, the method alleviates resource contention under high concurrency while preserving strong consistency semantics. Experimental evaluation demonstrates that the proposed technique significantly improves system throughput and response efficiency in high-load scenarios.
π Abstract
A significant impediment to high performance in key-value stores is the high cost of thread switching or stalls. While there are many sources for this, a major one is the contention for resources. And this cost increases with load as conflicting operations more frequently try to access data concurrently. Traditional latch-based approaches usually handle these situations by blocking one or more contending threads. Latch-free techniques can avoid this behavior. But the payoff may be limited if latch-free techniques require executing wasted work. In this paper, we show how latch-free techniques exploit delta record updating and can significantly reduce wasted work by using notices, a new latch-free approach. This paper explains how notices work and can solve B-tree index maintenance problems, while avoiding thread switches or stalls. Other opportunities for avoiding thread switches or stalls are also discussed.