# Dedicated Search Nodes two-node cost floor versus coupled mongot > Source-anchored research for Dedicated Search Nodes two-node cost floor versus coupled mongot. Run status: COMPLETED. Parent: [Cluster Tier Limits and Feature Gates as Cost Drivers (Connections, IOPS Caps, Minimum Tiers)](https://llms-explorer.com/tree/cluster-tier-limits-and-feature-gates-as-cost-drivers-connections-iops-caps-minimum-tiers/) · 5 facets · 11 facts · page: https://llms-explorer.com/tree/dedicated-search-nodes-two-node-cost-floor-versus-coupled-mongot/ ## How it works - The Atlas two-node minimum is not a universal mongot process requirement: Percona documents one mongot deployment serving a complete three-member replica set. This different product topology does not establish an Atlas single-node option or equivalent availability. (medium confidence) (verified as of 2026-10-29) — [source](https://docs.percona.com/percona-search-for-mongodb/vector-search-deployment.html) ## Measurements and reference values - For a simple deployment with exactly two equally priced Search Nodes, the inferred search-compute baseline is 2 × the selected per-node hourly rate × billed hours, on top of database charges. Network traffic between Search Nodes and database nodes adds charges; this formula is not a total-cost quote. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/search-node.md) ## Problems, failure modes and limitations - Dedicated Search Nodes require an M10-or-higher Atlas database cluster; Free and Flex clusters cannot add them. A small deployment moving to dedicated search must account for this database-tier eligibility gate as well as Search Node charges. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/cluster-config/multi-cloud-distribution.md) ## Comparisons and alternatives - Atlas enforces at least two dedicated Search Nodes for availability and charges hourly for each node; the minimum is a provisioned pair rather than a single optional search instance. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) ## Facts and statements - Atlas bills each Search Node hourly, including a full hour for partial-hour operation. Changing instance size incurs charges for both instances during the first hour. Traffic between Search Nodes and database nodes is chargeable. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/search-node.md) - Enabling coupled Atlas Search has no additional enablement fee, but mongot shares resources with mongod. This can be cheaper for small workloads while increased index resource use can affect database performance. (low confidence) (verified as of 2026-10-27) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) - Dedicated Search Nodes separate search compute from database compute and allow independent scaling. Their benefit is resource isolation; these sources do not establish a universal cost saving over coupled deployment. (high confidence) — [source](https://www.mongodb.com/docs/search/deployment/deployment-options.md) - MongoDB announced S10 on July 23, 2026 as a lower-cost tier below S20. Configuration documentation identifies S10 as the lowest tier, while S20 remains the default. S20 should therefore not be presented as the universal minimum tier. (high confidence) (verified as of 2026-07-23) — [source](https://www.mongodb.com/products/updates/introducing-the-s10-dedicated-search-node-tier/) - FHIRfly's August 5, 2026 incident account attributes cluster CPU usage to an unqueried autocomplete index over 9.6 million providers. This illustrates that coupled search capacity costs can persist despite negligible search-query volume; it does not quantify savings from dedicated nodes. (low confidence) — [source](https://fhirfly.io/blog/search-index-served-zero-queries) - A user reported overlapping coupled-Search failures during an October 12, 2026 scale-down, including mongot-is-shutting-down errors while database members remained reachable. This is anecdotal operational evidence, not a universal downtime duration or an SLA measurement. (low confidence) — [source](https://www.reddit.com/r/MongoDB_Official/comments/1x3wh0q/recurring_atlas_search_interruptions_during_both/) - For a single-region, unsharded Atlas deployment, the minimum dedicated search configuration is two billed Search Nodes, justified by high availability. Inference: if each node costs p per hour, the steady-state search compute floor is 2p per hour, in addition to the database cluster and applicable transfer charges. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/billing/search-node.md)