# Co-located vs dedicated search migration thresholds > Source-anchored research for Co-located vs dedicated search migration thresholds. Run status: COMPLETED. Parent: [Atlas Cluster Topology and Tier Economics](https://llms-explorer.com/tree/atlas-cluster-topology-and-tier-economics/) · 5 facets · 5 facts · page: https://llms-explorer.com/tree/co-located-vs-dedicated-search-migration-thresholds/ ## How it works - There is no blanket cost advantage to dedicated nodes. MongoDB says co-location can be more cost-efficient when existing cluster resources are sufficient; dedicated nodes add a separate billed resource, but can avoid scaling the entire database tier solely to serve search. Compare the incremental Search Node cost with the database-tier increase and isolation value for the measured workload. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) ## How-to and procedures - Use workload pressure as the production migration signal: compare Search index size with memory available to mongot, and monitor sustained query load, latency, ingestion and index-build pressure, and index growth. MongoDB sizing guidance treats index footprint, indexing load, QPS, and latency as separate sizing inputs. Practitioner guidance also describes moving when search affects OLTP performance or search demand drives database over-provisioning. (high confidence) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) ## Measurements and reference values - MongoDB's current MongoDB Search deployment guidance describes co-located nodes as suitable for testing and prototyping below all three limits: 2 million total documents to index, 10 GB of indexed data, and 10,000 queries in a 7-day period. It says to migrate to dedicated Search Nodes if usage exceeds the listed values. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) ## Problems, failure modes and limitations - Availability can justify migration before a workload shows steady-state contention. MongoDB documents that co-located Search may be unavailable during cluster scaling, node failover, or mongot upgrades and restarts. In a Reddit report, an operator described 55 final HTTP 503 responses during one scale-down test on a three-member cluster at about three Search requests per second; this is a single user report, not a general failure-rate estimate. (medium confidence) (verified as of 2026-10-22) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) ## Facts and statements - The numeric limits are scoped to testing and prototyping, not stated as a universal production cutoff. MongoDB separately recommends dedicated Search Nodes for production-ready MongoDB Search and Vector Search workloads, while its Vector Search guidance recommends co-location only for testing and prototyping. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md)