Egress and data transfer estimation
Parent: Atlas Cost Planning: Region Pricing, Estimation, Scheduled Scaling and Competitor TCO · Published reference · snapshot 2026-10-30
↓ Facts as markdownall context files
Source-anchored research for Egress and data transfer estimation. 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.
Problems, failure modes and limitations
- In its 2023 description, SERVER-79842 reports that serverStatus.network.compression counters combine application and compressed replication traffic. Treat this as a documented measurement limitation for that historical implementation, not proof of behavior in every current server version. (low confidence) [source]
- A March–April 2024 Atlas forum thread records unexpected transfer charges despite small stored datasets. MongoDB staff identify inefficient access patterns as a possible cause and request support investigation; the thread does not prove a root cause or billing error. (low confidence) [source]
Comparisons and alternatives
- Co-locating application traffic in the same provider and region, reducing returned fields and enabling driver network compression are supported ways to reduce transfer. Estimate savings from measured payloads and routes rather than assuming a universal compression percentage. (high confidence) [source]
- Atlas cloud-backup snapshot export adds $0.125/GB exported to object storage plus the cloud service's transfer costs. Atlas compresses exports; summing database dbStats.dataSize gives an uncompressed-size estimate and an upper bound for the Atlas export fee, rather than the complete cloud bill. (low confidence) (verified as of 2026-10-30) [source]
- Treat compression as a measured traffic-versus-compute tradeoff, not a guaranteed savings percentage. Atlas billing guidance recommends wire compression; MongoDB's July 8, 2026 synthetic benchmark varies read/write mix and compression settings and warns that higher zlib levels add CPU overhead. Its approximately 51 KB documents and M50 test deployment do not establish a universal production savings rate. (medium confidence) [source]
- Normalize billing units before comparing rates: Atlas labels its transfer tables per GB, while Google Cloud labels relevant VPC transfer tables per GiB. Confirm each bill's byte definition instead of treating these labels as interchangeable or importing Google's direct-customer allowances into an Atlas estimate. (medium confidence) (verified as of 2026-10-30) [source]
Facts and statements
- Estimate transfer by directed source-to-destination path and classify each path by provider, region, availability-zone boundary and internet/private route. Planning inference: sum each path's billable monthly GB multiplied by its applicable rate, then add route-specific processing and hourly charges. (high confidence) [source]
- For an initial application-read volume estimate, multiply average returned document size by documents served per second and by active seconds in the month. This is a payload baseline: measure actual compressed traffic and map it to billing paths before treating it as billable GB. (high confidence) [source]
- Stored database size alone is insufficient for estimating egress: repeated reads, response size, request frequency and the boundaries crossed determine transferred volume. Returning only required fields and avoiding repeated reads reduce that volume. (high confidence) [source]
- Build separate traffic budgets for application responses, inter-region replication, analytics/export jobs and observability traffic. A multi-region topology adds ongoing replication paths; include these paths before comparing total costs with a single-region topology. (high confidence) [source]
- Include customer-side networking charges in Atlas TCO. The beSharp public-endpoint example includes a NAT gateway hourly charge and per-GB processing, whereas its peering example replaces that route. Its conclusion depends on the assumed architecture and rates. (high confidence) [source]
- AWS recommends validating transfer estimates using a proof of concept with realistic simulated load, then comparing different demand levels. For infrastructure in your AWS account, Cost Explorer or the Cost and Usage Report can inform the model; this does not imply visibility into Atlas-managed infrastructure charges. (high confidence) [source]
- Do not remove cross-zone or cross-region redundancy solely to eliminate transfer fees. Model its cost against the reliability outcome it provides and compare architectural alternatives at the same required resilience. (high confidence) [source]
- The cached Atlas transfer documentation states that incoming transfers are not charged by Atlas, and its AWS and Google Cloud sections exempt Free and Flex clusters from outgoing-transfer charges. These exceptions do not establish that the source system or customer networking path is free. (low confidence) (verified as of 2026-10-14) [source]
- Atlas tabulates transfer costs daily and varies pricing by cloud provider and source/destination region. Its AWS table lists $0.00–$0.01/GB for same-region transfer, $0.02–$0.15/GB for different-region transfer, and $0.09–$0.16/GB for internet egress. These are published ranges, not a quote for a particular deployment. (low confidence) (verified as of 2026-10-30) [source]
- Model transfer by source, destination, direction, volume, and frequency; evaluate low, expected, and high usage scenarios. A planning calculation is the sum of billable volume times the applicable rate for each path, adjusted for that tariff's units, tiers, and exceptions. This is a modeling synthesis, not an Atlas calculator output. (medium confidence) [source]
- Do not assume same-region placement eliminates transfer charges. Atlas publishes nonzero same-region categories; Google Cloud separately distinguishes same-zone internal traffic, inter-zone traffic, and external-address paths. Use Atlas rates for Atlas billing and provider rates only for the separately applicable provider charges. (medium confidence) (verified as of 2026-10-30) [source]
- The AWS and Google Cloud sections of the Atlas billing page exempt incoming transfers and outgoing transfers from Free/Flex clusters. The page separately warns that Azure virtual-network traffic can incur ingress and egress charges. Scope exemptions to the relevant provider and billing surface; free ingress at a destination does not establish a free end-to-end migration. (low confidence) (verified as of 2026-10-30) [source]
- Account for replication and non-application traffic. Atlas warns that additional regions can increase transfer costs, platform operations contribute background traffic, and external-service log exports route through the control plane in AWS us-east-1, potentially adding cross-region or inter-cloud transfer alongside internet egress. (low confidence) (verified as of 2026-10-30) [source]
- Atlas recommends reducing returned fields with projection, avoiding repeated reads of data already held by the client, locating applications in the same provider/region where practical, and enabling client-driver network compression. Its documentation states that intra-cluster communication is already compressed. (low confidence) (verified as of 2026-10-30) [source]
Children
- Traffic-path and monthly-volume modeling (frontier)
- Compressed wire-volume measurement (frontier)
- Transfer billing reconciliation (frontier)
- Application payload reduction (frontier)
Frontier under this node: Application payload reduction, Compressed wire-volume measurement, Traffic-path and monthly-volume modeling, Transfer billing reconciliation