AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Home Made CobbleDB Replaces DynamoDB at Perplexity to Cut Query Latency 5x and Reduce Cloud Storage

Collected Sep 30, 2026

Perplexity has moved its core search serving tier from Amazon DynamoDB to CobbleDB, an internally developed distributed key-value store written in Rust, according to InfoQ. The migration addressed latency and cost bottlenecks arising from serving multi-kilobyte document batches to large language models under heavy query volumes.

Each Perplexity query generates between 100 and 120 target page keys, which the retrieval service splits into parallel batches of 10 to 20 keys, producing average record payloads of roughly 50 kilobytes. At production traffic exceeding 200,000 requests per second, DynamoDB usage-based pricing became unsustainable because AWS meters every byte transferred, and the team could not prevent tail-latency spikes from uncached reads, cross-zone networking hops, or lagging replicas.

Perplexity split its storage architecture into three systems: Pillar for durable state management on YTsaurus, Lorry for stateless batch aggregation that stores payloads in Amazon S3, and CobbleDB for low-latency serving. CobbleDB uses RocksDB as its embedded storage engine with memory-mapped caching and local NVMe disks, maintains three replicas per partition, and routes reads to same-availability-zone replicas with speculative hedging. It discards distributed transaction protocols and synchronous consensus, applying updates asynchronously.

In live production measurements, CobbleDB reduced median batch-read latency from 31.4 milliseconds to 5.60 milliseconds, p90 from 56.7 milliseconds to 9.77 milliseconds, and p99 from 123 milliseconds to 24.2 milliseconds. Synthetic benchmarks handling payloads up to 100 kilobytes confirmed throughput up to 500,000 requests per second.

The architecture transfers node lifecycle management, backup verification, and partition rebalancing to internal site reliability engineers, and applications must withstand eventual consistency. CEO Aravind Srinivas said the 40,000 lines of Rust were built in two months by two systems engineers paired with an autonomous swarm of AI coding agents. Perplexity plans to open-source CobbleDB.

Read at InfoQ · AI, ML & Data Engineering

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Perplexity has migrated its search infrastructure from Amazon DynamoDB to CobbleDB, an internally developed key-value store in Rust. This change reduced latency and costs associated with handling large document batches. The new architecture supports high query volumes more efficiently, achieving improved latency and reduced storage expenses while managing significant production traffic. By Olimpiu Pop