Search Node network transfer and multi-region cost
Parent: Platform Consolidation TCO (Atlas Search/Vector Search vs Separate Search or Vector DB) · Published reference · snapshot 2026-10-30
↓ Facts as markdownall context files
Source-anchored research for Search Node network transfer and multi-region cost. Run status: COMPLETED.
These notes link each claim to its source. Some citations identify archived research reports that are no longer hosted on this site. A published reference means the content is available; it does not certify independent review or accuracy.Read the editorial policy and follow the sources before relying on a claim.
How it works
- MongoDB's 2024 engineering guidance uses regional and availability-zone replica-set tags with secondary read preferences to improve locality, with fallback when preferred nodes are unavailable. This is conditional secondary-read guidance, not proof that all search or write traffic becomes local. (low confidence) (verified as of 2024-08-30) [source]
- Atlas recommends placing applications in the same provider and region, using nearest read preference when cross-region reads are necessary, limiting returned fields, and enabling client network compression. Atlas also states that intra-cluster communication is always compressed. These are optimizations, not evidence that all search traffic becomes free. (low confidence) (verified as of 2026-10-14) [source]
Comparisons and alternatives
- MongoDB's multi-region Search Node announcement describes independent scaling for both Atlas Search and Vector Search, plus regional redundancy. A consolidation comparison should price equivalent geographic availability on both sides rather than comparing one region against several. (medium confidence) (verified as of 2025-03-06) [source]
- For a TCO comparison, measure the actual transfer volume on each route and apply the relevant tariff separately from Search Node capacity. This is an inference from Atlas internal-traffic billing and cloud-provider route-dependent charges; the sources provide no universal network-savings percentage for Atlas versus a separate search/vector service. (medium confidence) (verified as of 2026-10-14) [source]
Facts and statements
- Atlas bills network traffic between dedicated Search Nodes and database nodes. These charges appear in cluster-level data transfer, without a separate Search Node transfer breakdown; consolidation therefore does not imply zero internal search-network cost. (low confidence) (verified as of 2026-10-14) [source]
- Atlas transfer charges depend on provider and source/destination regions, and multi-region clusters can incur higher transfer costs. Model application traffic and internal Search Node traffic separately before aggregating the cluster bill. (medium confidence) (verified as of 2026-10-14) [source]
- Atlas requires at least two Search Nodes per region and supports counts per region per shard. Its multi-region configuration uses the same search tier across regions and deploys the specified count in each region containing database nodes for sharded clusters; regional replication can thus multiply search compute as well as transfer exposure. (medium confidence) (verified as of 2026-10-14) [source]
- Same-region placement does not guarantee free networking: AWS charges applicable peering transfers while exempting same-AZ peering, and Google Cloud distinguishes free qualifying same-zone VM traffic from charged cross-zone traffic. Provider tariffs explain topology effects but are not substitutes for Atlas invoice rates. (medium confidence) (verified as of 2026-10-30) [source]
- Google Cloud's VM tariff lists $0.01/GiB for different zones in one region, $0.02/GiB between North American regions, and $0.05/GiB between North America and Europe. These are cloud-VM comparison inputs, not quoted Atlas Search Node prices. (low confidence) (verified as of 2026-10-30) [source]
- AWS peering has no connection-creation fee but can incur transfer fees. A beSharp Atlas connectivity example shows that replacing a NAT-routed public path with peering removes NAT processing from that path while leaving possible cross-AZ charges; the saving depends on the actual route and existing infrastructure. (medium confidence) (verified as of 2026-01-14) [source]
- MongoDB recommends same-provider, same-region application placement, smaller projected result payloads, and client-driver compression to reduce transferred bytes. Its cost guidance states that intra-cluster communication is already compressed; client compression should not be credited again as a new internal Search Node saving. (medium confidence) (verified as of 2026-10-30) [source]
- A 2020 Atlas forum complaint attributed network-out charges to search indexing, while a MongoDB reply described search reading only the local mongod. The thread establishes a historical attribution dispute, not a demonstrated billing defect or a current exemption for dedicated Search Nodes. (medium confidence) [source]
- For TCO, estimate each retained network path as transferred bytes times the applicable direction/location rate, then include relevant gateway charges and regional Search Node hours. This is a modeling inference from Atlas billing and a worked connectivity example; the sources do not establish a universal consolidation savings percentage. (medium confidence) [source]
- Consolidating search into Atlas does not establish a zero-transfer-cost architecture: dedicated Search Nodes have billable internal traffic, and Atlas transfer pricing depends on the cloud provider and source/destination regions. (high confidence) (verified as of 2026-10-14) [source]
- Atlas requires at least two Search Nodes per region and charges hourly for each Search Node. For sharded multi-region clusters, it deploys the configured Search Node count in each region containing database nodes; regional expansion therefore adds a capacity cost as well as possible transfer costs. (high confidence) (verified as of 2026-10-14) [source]
- MongoDB announced multi-region Search Nodes across AWS, Google Cloud and Azure. The configuration documentation scopes that availability: Search Nodes deploy in regions containing electable, read-only or analytics nodes, with unsupported-region exclusions. The announcement does not mean every cloud region supports them. (medium confidence) (verified as of 2026-10-14) [source]
- Same-region placement does not universally make network traffic free. AWS charges VPC peering traffic crossing availability zones, while Google prices eligible same-region inter-zone VM traffic at $0.01 per GiB. These cloud-provider rules describe connectivity costs, not a substitute tariff for the Atlas invoice. (high confidence) (verified as of 2026-10-30) [source]
- AWS says creating a VPC peering connection is free, and peered traffic remaining within one availability zone is free even across accounts; crossing availability zones or AWS regions incurs transfer charges. Private peering alone is therefore insufficient evidence of zero network cost. (low confidence) (verified as of 2026-03-19) [source]
- Google lists no charge for eligible same-zone VM traffic using internal IPv4 or IPv6 addresses. Its separate inter-region VM transfer table varies by source and destination geography, and external-address paths can use internet rates. Cost estimates must retain the actual route and GiB billing unit. (low confidence) (verified as of 2026-10-30) [source]
Children
- Search Node traffic attribution and measurement (frontier)
- Regional Search Node topology cost modeling (frontier)
- Private networking gateway and cross-zone fees (frontier)
Frontier under this node: Private networking gateway and cross-zone fees, Regional Search Node topology cost modeling, Search Node traffic attribution and measurement