Write-Read Decoupling in Modern Large-Scale Search Engines: Architectures, Techniques, and Emerging Approaches

TL;DR

This survey analyzes architectural solutions for decoupling write and read operations in large-scale search engines to reduce latency and resource contention, highlighting emerging hybrid approaches like ScaleSearch.

Contribution

It systematically reviews existing architectures, introduces the ScaleSearch synthesis, and discusses open challenges in hybrid retrieval, AI integration, and serverless deployment.

Findings

Identifies five principal patterns for write-read decoupling.

Highlights the ScaleSearch architecture combining multiple techniques.

Discusses open challenges in hybrid retrieval and AI integration.

Abstract

Large-scale search engines face a fundamental tension: the index must be updated frequently to maintain freshness, yet updates create resource contention that inflates query latency. In the dominant Lucene-based architecture, segment merges triggered by writes compete with concurrent queries for CPU cycles, disk I/O bandwidth, and operating-system page cache -- a problem we term \emph{write-read contention}. This survey systematically examines the architectural solutions that industry and academia have developed to decouple write pressure from read latency. We identify five principal patterns: (i)~node-level read-write separation; (ii)~compute-storage separation; (iii)~full in-memory indexing; (iv)~log-structured write paths; and (v)~in-place partial updates. We survey representative systems including Elasticsearch, LinkedIn Galene, Uber Sia, Quickwit, Alibaba Havenask, Algolia, Milvus, and Vespa, and discuss an emerging synthesis -- the ScaleSearch architecture -- that combines compute-storage separation with full in-memory indexing and dedicated write nodes. A key contribution of ScaleSearch is \emph{per-field update routing}: each field is assigned its own Kafka topic and update path, allowing scalar fields (price, stock, tags) to be updated in-place in $O(1)$ RAM with immediate visibility while full-text fields follow the segment-based compute-storage path. We conclude with open challenges in hybrid vector-and-full-text retrieval, serverless deployments, and AI-integrated search.