Online Archive Data Access and Data Seek Query Cost Model
Parent: Online Archive partition-field selection · Published reference · snapshot 2026-10-30
↓ Facts as markdownall context files
Source-anchored research for Online Archive Data Access and Data Seek Query Cost Model. 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.
Structure and components
- The published processing charge is $5 per TB with a 10 MB minimum per query. Data Federation rounds processed bytes up to the nearest megabyte. This is a processing component, not the complete Online Archive bill. (medium confidence) (verified as of 2026-10-14) [source]
How it works
- The September 29, 2026 pricing table lists AWS Northern Virginia archival access at $0.001 per 1000 partitions accessed. Google Cloud Iowa lists $0.0008 per 1000 requests; preserve this difference in units when estimating costs. (low confidence) (verified as of 2026-09-29) [source]
- A Reddit pricing discussion reports difficulty predicting query expense because reads depend on the stored data accessed. This is practitioner testimony, not a benchmark or verified invoice. (low confidence) [source]
How-to and procedures
- Use the federated endpoint to inspect archive query plans. Data Federation explain exposes stats.size and stats.numberOfPartitions for partitions it might open, and Online Archive optimizationPlan describes archive optimizations. These are planning diagnostics; they should not be presented as an invoice. (medium confidence) [source]
Comparisons and alternatives
- Treat frequent historical scans as a cost-risk workload. TideWave warns that hourly heavy analytics can accumulate query charges; Harrison describes the storage-versus-retrieval trade-off. Savings from cheaper cold storage require evaluation alongside actual query frequency and pruning. (medium confidence) [source]
Facts and statements
- Data Access measures archive partitions downloaded for a query: MongoDB describes each downloaded partition as one access. The AWS and Azure pricing tables express archival access per 1,000 partitions accessed; this unit is separate from processed bytes. (medium confidence) (verified as of 2026-10-14) [source]
- Data Seek describes operations that discover the partitions needed by a query. Each operation finds up to 1,000 partitions, and Atlas performs the minimum required operations; the documentation example finds 100 query-covered partitions with one operation. This does not mean those 100 downloads count as one Data Access. (low confidence) [source]
- Both federated and archive-only queries incur scan, access, seek, and transfer costs; selecting the archive-only endpoint does not eliminate query charges. (low confidence) [source]
- Federated scans push eligible work to the live cluster. A sort without a match retrieves all cluster and archive data. (low confidence) [source]
- Estimate monthly processing from representative query work and monthly execution frequency, accounting for the processing minimum. MongoDB recommends explain-based estimation, configurable processing limits, and billing alerts; the billing page provides actual costs while the federation page reports usage. (low confidence) [source]
- Tidewave recommends choosing partition fields from actual historical-query filters and warns that frequent heavy archive analytics can accumulate query charges. This is qualitative guidance, without a measured break-even point. (low confidence) [source]
- A MongoDB community response describes partition indexes and a managed storage service beyond S3 alone. Its architecture explanation cautions against modeling the service solely as an S3 storage class. (low confidence) [source]
- Storage, archival access, processing, and transferred or returned data are separate cost dimensions. Storage rates depend on provider and region, so estimating query processing alone omits other billable usage. (medium confidence) (verified as of 2026-10-14) [source]
- Partition filters can reduce the archive partitions read. Guy Harrison illustrates customer-specific partitioning that avoids scanning other customers, and TideWave likewise connects partition pruning to lower query work. This supports choosing partition fields from actual historical-query filters. (medium confidence) [source]
- Partitioning alone does not guarantee lower processing cost: Data Federation explicitly says an empty $match still reads the entire collection. A query must actually exclude data to gain pruning savings. (low confidence) [source]
- Data Federation bills transferred and returned bytes separately from processed data. Its transfer total includes client results, transfers between query nodes, and writes through $out or $merge; transfer pricing depends on provider and network path. A small result therefore does not establish a small processing charge. (medium confidence) [source]
Children
- No children recorded.