Serializable HTAP with Abort-/Wait-free Snapshot Read

TL;DR

This paper introduces a novel multiversion concurrency control method called read safe snapshot (RSS) that achieves serializability in HTAP systems without additional aborts or waits, enhancing performance for OLTP and OLAP coexistence.

Contribution

The paper proposes a new RSS method based on MVCC and SSI, enabling serializable HTAP transactions with fewer aborts and waits, and demonstrates its practical implementation in open-source databases.

Findings

RSS reduces abort rates in HTAP systems.

The method improves performance in single-node architectures.

Multi-node architecture evaluation shows acceptable overhead.

Abstract

Concurrency Control (CC) ensuring consistency of updated data is an essential element of OLTP systems. Recently, hybrid transactional/analytical processing (HTAP) systems developed for executing OLTP and OLAP have attracted much attention. The OLAP side CC domain has been isolated from OLTP's CC and in many cases has been achieved by snapshot isolation (SI) to establish HTAP systems. Although higher isolation level is ideal, considering OLAP read-only transactions in the context of OLTP scheduling achieving serializability forces aborts/waits and would be a potential performance problem. Furthermore, executing OLAP without affecting OLTP as much as possible is needed for HTAP systems. The aim of this study was serializability without additional aborts/waits. We propose read safe snapshot (RSS) using multiversion CC (MVCC) theory and introduce the RSS construction algorithm utilizing serializable snapshot isolation (SSI). For serializability of HTAP systems, our model makes use of multiversion and allows more schedules with read operations whose corresponding write operations do not participate in the dependency cycles. Furthermore, we implemented the algorithm practically in an open-source database system that offers SSI. Our algorithm was integrated into two types of architecture as HTAP systems called as unified (single-node) or decoupled (multinode) storage architecture. We evaluate the performance and abort rate of the single-node architecture where SSI is applicable. The multi-node architecture was investigated for examining the performance overhead applying our algorithm.