TECHNOLOGY PROFILE · TECH-0287
Vespa
Search and recommendation serving engine with programmable ranking.
Reference link only — not reviewed this edition. This profile preserves the prepared catalog’s original descriptions. Advantages depend on the workload; comparisons are editorial, not benchmark results.
Where it has an edge
Combines retrieval and multi-stage ranking close to indexed data rather than shipping candidates through many services.
Best-fit problems
Complex hybrid search, recommendations and feature-rich ranking.
Tradeoffs & cautions
Schema, ranking and cluster design have a steeper learning curve.
What to test before choosing
Measure recall@k, filtered recall, indexing cost, p95 query latency, updates, backup and access control.
Runtime & hardware
Embedded, self-hosted or managed
Access & licensing
Open-source or commercial depending on edition
Execution
Local / self-hosted
Lifecycle
Not independently audited
Comparison basis
Editorial problem-fit judgement; no head-to-head benchmark performed
Connected concepts
Alternatives to investigate
Elasticsearch · OpenSearch · Milvus