# Elasticsearch-to-Atlas Search migration and dual-run cost > Source-anchored research for Elasticsearch-to-Atlas Search migration and dual-run cost. Run status: COMPLETED. Parent: [Platform Consolidation TCO (Atlas Search/Vector Search vs Separate Search or Vector DB)](https://llms-explorer.com/tree/platform-consolidation-tco-atlas-search-vector-search-vs-separate-search-or-vector-db/) · 4 facets · 14 facts · page: https://llms-explorer.com/tree/elasticsearch-to-atlas-search-migration-and-dual-run-cost/ ## Measurements and reference values - MongoDB reports 30–50% improved development velocity and approximately 10% of engineering cycles spent recovering synchronization failures, attributing the latter to interviews with teams that replaced external search. These are vendor-reported observations without a published sampling methodology on the page; neither figure is a general infrastructure savings percentage. (low confidence) — [source](https://www.mongodb.com/company/blog/technical/5-steps-replacing-elasticsearch-solr-atlas-search) ## Problems, failure modes and limitations - Meltwater's 2022 Elasticsearch-version migration kept two clusters live to preserve rollback; introducing the second cluster created update-order and consistency problems. This is adjacent search-migration evidence for dual-run engineering risk, not an Elasticsearch-to-Atlas case or an Atlas cost benchmark. (low confidence) — [source](https://underthehood.meltwater.com/blog/2022/11/18/how-we-upgraded-an-old-3pb-large-elasticsearch-cluster-without-downtime-part-2-two-consistent-clusters/) - A Reddit commenter reports migrating hundreds of Elasticsearch nodes to Atlas Search, testing first, retaining a rollback plan, and using fewer search nodes afterward. This anonymous, unverified account provides a follow-up lead; it supplies neither a dollar denominator nor overlap duration and cannot establish general savings. (low confidence) — [source](https://www.reddit.com/r/mongodb/comments/1n2f3hg) - In its November 2022 Elasticsearch-to-Elasticsearch upgrade account, Meltwater reported a team of four people spending over a year redesigning the indexing/update pipeline and approximately 110 days filling the new cluster. This unusually large mutable-data workload is a risk example, not an estimate for an Elasticsearch-to-Atlas Search migration. (low confidence) — [source](https://underthehood.meltwater.com/blog/2022/11/18/how-we-upgraded-an-old-3pb-large-elasticsearch-cluster-without-downtime-part-2-two-consistent-clusters/) - Gravity9 reports an Atlas Search migration framework with query conversion and support for legacy and new search methods at a financial institution. The case does not identify the legacy engine or publish an overlap duration, migration bill, or realized savings amount, so it cannot substantiate a quantified Elasticsearch replacement ROI. (low confidence) — [source](https://www.gravity9.com/case-studies/global-financial-institution-case-study/) ## Comparisons and alternatives - If MongoDB already holds the authoritative application data, Atlas Search can synchronize from that data rather than requiring a separately maintained Elasticsearch synchronization pipeline. The retired pipeline is a potential recurring operations saving, not proof of a particular dollar saving. (high confidence) — [source](https://www.mongodb.com/company/blog/technical/5-steps-replacing-elasticsearch-solr-atlas-search) - For application data already stored in MongoDB Atlas, the consolidation opportunity includes retiring the separately managed database-to-Elasticsearch synchronization pipeline. If the authoritative database remains outside MongoDB, an ingestion path into MongoDB is still needed; gravity9 describes redirecting the existing pipeline rather than eliminating it. (medium confidence) — [source](https://www.mongodb.com/company/blog/technical/5-steps-replacing-elasticsearch-solr-atlas-search) ## Facts and statements - An Elasticsearch-to-Atlas Search migration includes workload qualification, index mapping, query rewrites, and relevance/performance validation; consolidating the platform does not remove this application migration work. (high confidence) — [source](https://www.mongodb.com/company/blog/technical/5-steps-replacing-elasticsearch-solr-atlas-search) - A parallel migration trades safer testing and rollback for temporary duplication of infrastructure and synchronization work. Elastic describes these costs for Elasticsearch cluster migrations, and Meltwater documents them in production; this supports budgeting an overlap phase for an Atlas migration, but does not establish that every Atlas migration requires dual writes. (medium confidence) — [source](https://www.elastic.co/search-labs/blog/elasticsearch-migrate-data-versions-clusters) - Elastic identifies temporary hardware duplication, configuration and testing of the second cluster, and synchronization of writes during parallel migration. Pureinsights offers a fixed-price migration engagement with a tailored Statement of Work. These identify distinct infrastructure and migration-service costs to obtain for a workload-specific estimate; they do not provide a universal total-spend multiplier. (medium confidence) — [source](https://underthehood.meltwater.com/blog/2022/11/18/how-we-upgraded-an-old-3pb-large-elasticsearch-cluster-without-downtime-part-2-two-consistent-clusters/) - Pureinsights' migration-services brief recommends a search assessment for complex migrations and identifies ongoing relevance tuning after migration. A TCO estimate should therefore include assessment and continuing search-quality work even after consolidation. (low confidence) — [source](https://pureinsights.com/wp-content/uploads/2023/08/Solution-Brief-Atlas-Search-Migration-Services-V4.pdf) - Gravity9’s September 2022 migration advice scopes Atlas Search primarily to application search and warns that complex observability migrations can require enough additional implementation to outweigh the benefits. Treat this as historical workload-fit guidance and requalify current features before assuming the entire Elasticsearch estate can be retired. (low confidence) — [source](https://www.gravity9.com/blog/a-spotlight-on-migrating-elasticsearch-to-atlas-search/) - During a live backfill, account for concurrent additions, updates, record identity, and update ordering. Elastic discusses replay and duplicate prevention; Meltwater had to preserve effective update ordering across both clusters. These are migration consistency concerns, not evidence of an Atlas-specific implementation requirement. (high confidence) — [source](https://www.elastic.co/search-labs/blog/elasticsearch-migrate-data-versions-clusters) - MongoDB's migration guide includes workload qualification, index migration, query migration, query validation, and sizing/deployment. Meltwater's adjacent Elasticsearch upgrade required a team of four people working for over a year and about 110 days to fill the target cluster. These sources demonstrate that transition work and overlap duration depend on the migration; they do not give a universal payback period for an Atlas Search migration. (medium confidence) — [source](https://www.mongodb.com/company/blog/technical/5-steps-replacing-elasticsearch-solr-atlas-search)