The Faster, The Cheaper
summary
Orama is a distributed full‑text and vector search engine that proves a claim: immutability at the edge makes systems cheaper and faster as load rises. Implementation uses JavaScript + WASM (Rust for hot code), AVL trees and structural sharing, indexes deployed as immutable shards to a CDN (≈310 global locations), a Cortex cache layer and KV storage. Traffic jumps from ~300k to 6–8M requests/month with peaks of 1,400 rps, 100 ms average RTT, and significant CPU time reduction as caches warm, showing the approach works.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 2:54Why Orama chose JavaScript and WebAssembly
- 7:30Orama’s scalable architecture
- 10:16Functional programming meets distributed systems
- 13:50Immutable data dilemma
- 15:14How Orama uses CDNs for global distribution
- 18:27Performance numbers
- 21:54Key lessons
- 22:36Q&A
related talks
19mDisrupting Full Text Search Industry With JavaScript
41mScaling Search and Powering Vector-Based Retrieval With OpenSea...
28mSearch and Sushi; Freshness Counts
45mAccelerating OpenSearch With Streaming: Apache Arrow, Flight, Data...
01h 03mExtreme Performance Architecture
28mLyra: Disrupting the full-text search industry with JS
47mRust At The Edge
39mSuper Fast Real-time Data Processing on Cloud-Native Architecture