[Paper Review] NWR: Rethinking Thomas Write Rule for Omittable Write Operations

Sho Nakazono, Hiroyuki Uchiyama+3 more|arXiv (Cornell University)|Apr 17, 2019

Distributed systems and fault tolerance39 references4 citations

TL;DR

This paper introduces the Non-visible Write Rule (NWR), a formal framework that enables any concurrency control protocol to safely omit write operations without locks, by validating an additional version order. NWR decouples omittability from timestamp ordering, allowing protocols like Silo, TicToc, and MVTO to achieve up to 11x performance improvement on highly write-contended workloads while preserving strict serializability and recoverability.

ABSTRACT

Concurrency control protocols are the key to scaling current DBMS performances. They efficiently interleave read and write operations in transactions, but occasionally they restrict concurrency by using coordination such as exclusive lockings. Although exclusive lockings ensure the correctness of DBMS, it incurs serious performance penalties on multi-core environments. In particular, existing protocols generally suffer from emerging highly write contended workloads, since they use innumerable lockings for write operations. In this paper, we rethink the Thomas write rule (TWR), which allows the timestamp ordering (T/O) protocol to omit write operations without any lockings. We formalize the notion of omitting and decouple it from the T/O protocol implementation, in order to define a new rule named non-visible write rule (NWR). When the rules of NWR are satisfied, any protocol can in theory generate omittable write operations with preserving the correctness without any lockings. In the experiments, we implement three NWR-extended protocols: Silo+NWR, TicToc+NWR, and MVTO+NWR. Experimental results demonstrate the efficiency and the low-overhead property of the extended protocols. We confirm that NWR-extended protocols achieve more than 11x faster than the originals in the best case of highly write contended YCSB-A and comparable performance with the originals in the other workloads.

Motivation & Objective

• To address the performance bottleneck caused by excessive locking in write-contended workloads on multi-core systems.
• To formalize the concept of 'omittable' write operations independently of timestamp ordering, moving beyond the limitations of the classical Thomas Write Rule (TWR).
• To enable any concurrency control protocol to safely omit write operations by validating an additional version order, preserving strict serializability and recoverability.
• To design and implement NWR-extended protocols that achieve high performance with minimal overhead on modern benchmarks.

Proposed method

• Formalize 'omittability' using version orders from multiversion serializability theory, defining that a write is omitable if a reverse version order exists.
• Decouple omittability from timestamp-based protocols by introducing the Non-visible Write Rule (NWR), which relies on validating a reverse version order.
• Extend existing protocols (Silo, TicToc, MVTO) with NWR by adding validation of an additional version order to detect and omit obsolete writes.
• Use a variant of MVSG (Multi-Version Serializability Graph) validation to check consistency, with optimizations to reduce cache overhead by reusing graph structures.
• Implement a mechanism to detect blind updates and apply NWR only when the write is non-visible and can be safely omitted.
• Optimize traversal in version lists by reusing existing data structures and minimizing redundant operations, especially in MVTO+NWR.

Experimental results

Research questions

• RQ1 Can the concept of omittable write operations be generalized beyond timestamp-ordered protocols like T/O?
• RQ2 What formal conditions must be satisfied to safely omit a write operation without sacrificing strict serializability or recoverability?
• RQ3 Can a protocol be extended to validate multiple version orders efficiently, especially in high-contention workloads?
• RQ4 How does the performance of NWR-extended protocols compare to their original counterparts across diverse workloads?

Key findings

• NWR-extended protocols achieve more than 11x performance improvement over the original Silo in the highly write-contended YCSB-A benchmark.
• In the TPC-C benchmark, which features high read and write contention, NWR-extended protocols maintain performance comparable to the original protocols.
• The MVTO+NWR variant incurs additional overhead due to double traversal of version lists, but this is mitigated in practice by the efficiency of the reverse version order validation.
• The approach of validating a single additional version order enables efficient and correct write omission without requiring centralized transaction management or complex coordination.
• The PV-Rule of NWR ensures that aborts only require validation of a connected subgraph, reducing the computational cost of maintaining multiple version orders.
• The NWR framework is universally applicable: any protocol can be extended with NWR by adding a single reverse version order check, enabling safe write omission across diverse concurrency control models.