Atlas Online Archive: Cost & Usage

#atlas #mongodb #online-archive

TL;DR; (Quick Read)

  • MongoDB Atlas Online Archive moves “cold” data out of your primary cluster into lower-cost object storage while keeping it queryable: https://www.mongodb.com/docs/atlas/online-archive/
  • This estimate models 10 TB of archived data and provides annual costs for US East (N. Virginia) and South America (São Paulo).
  • Total cost = Storage + Query Processing + (potential) Network Egress
    • Storage is the predictable baseline (GB-days). For 10 TB it’s roughly:
      • ~$5.9K/year (US East)
      • ~$10.5K/year (São Paulo)
    • Query Processing is billed at $5 per TB processed (i.e., TB scanned), not “per query.” Well-filtered, date-bounded queries keep this low:
      • Light: ~$120/year (2 TB processed/month)
      • Moderate: ~$1,200/year (20 TB processed/month)
      • Heavy: ~$12,000/year (200 TB processed/month)
  • Network egress is a cloud-provider charge (not MongoDB). It’s usually minimal when reporting runs in the same cloud region as Atlas, and it becomes material when results are pulled to on-prem, exported in bulk, or moved cross-region: https://www.mongodb.com/docs/atlas/billing/data-transfer-costs/
  • The tables use Small/Medium/Large egress ranges to illustrate this uncertainty. The Atlas billing dashboard + invoice are the source of truth for final costs.

Introduction

This document provides a practical, experience-based overview of MongoDB Atlas Online Archive and a set of cost estimates for archiving approximately 10 TB of data.

The goal is to:

  • Explain how Online Archive pricing works in plain terms
  • Provide storage, query, and egress estimates for two regions
  • Show how costs change under Light / Moderate / Heavy usage patterns
  • Call out where costs are predictable vs. usage-driven
  • Highlight scenarios where network egress may apply

All numbers below are estimates intended for planning and comparison purposes. Final costs depend on actual usage patterns.

What is MongoDB Atlas Online Archive?

Atlas Online Archive allows you to transparently move infrequently accessed (“cold”) data from your primary Atlas cluster into low-cost cloud object storage, while still keeping the data queryable using standard MongoDB queries. See: Online Archive Overview.

From an application or reporting perspective:

  • Hot data stays in the Atlas cluster
  • Archived data lives in object storage
  • Queries can span both without changing application logic

Online Archive is commonly used for:

  • Reporting and analytics on historical data
  • Compliance and audit lookbacks
  • Cost optimization for large datasets

Pricing Components (What You Pay For)

Online Archive pricing has three main components (details: Online Archive Billing):

1. Archive Storage (GB-days)

  • Charged based on how much data is stored and how long
  • Billed as $/GB-day, varies by region
  • This is the most predictable cost

2. Query Processing Rate

Query Processing Rate is the cost charged when queries run against archived data (implemented via Atlas Data Federation; cost model described in Data Federation Billing).

Definition (plain English):
You are billed based on how much archived data MongoDB has to read/scan to answer a query — not how much data is returned.

Example:

  • Query scans 200 GB of archived data → 0.2 TB × $5 = $1
  • Query scans 10 TB of archived data → $50

Key control lever:
Well-filtered, date-bounded queries keep this cost very low. Broad historical scans increase it linearly.

3. Network Egress (Cloud Provider Charge)

Network egress is not a MongoDB charge.
It is charged by the underlying cloud provider when data leaves the Atlas cloud boundary.

Egress generally applies when:

  • Data is returned to on-prem systems
  • Data is exported to files (CSV, Parquet, etc.)
  • Queries cross regions or clouds
  • Large result sets are downloaded externally

If reporting tools run in the same cloud and region, egress is often minimal or near zero.

Assumptions Used in This Estimate

  • Total archived data: 10 TB (10,240 GB)
  • Archive type: Standard (non-time-series)
  • Pricing modeled for:
    • US East (N. Virginia)
    • South America (São Paulo)
  • Query Processing tiers (TB processed per month):
    • Light: 2 TB / month
    • Moderate: 20 TB / month
    • Heavy: 200 TB / month
  • Egress modeled as annual ranges due to cloud/provider variability

How to read the annual cost tables

All columns in the tables are annual USD cost estimates for a 10 TB Online Archive deployment, with usage modeled at different levels. The tables show costs broken into:

  1. Storage Cost (Annual USD)
  2. Query Processing Cost (Annual USD)
  3. Network Egress (Annual USD, Range) (provider-dependent, so shown as a range)
  4. Total Annual Cost Estimate (USD, Range)

Universal assumptions used in calculations

  • Archive size: 10 TB = 10,240 GB
  • Annualization: 365 days/year and 12 months/year
  • Costs shown are estimates for planning; actual invoices vary by usage, provider, destination, and transfer patterns (see Online Archive Billing and Atlas Data Transfer Costs).

Important: These figures are directional estimates based on published unit rates and assumed usage patterns. Actual billed amounts depend on configured archive criteria, data shape/compression, query selectivity, and the cloud provider’s networking rates and routing (including private connectivity vs. public internet). The invoice and Atlas billing dashboard are the source of truth.

1) Storage Cost (Annual USD)

What the number represents

Annual cost to store 10TB in Online Archive (regardless of queries).

Unit in table header

Storage Cost (Annual USD)

How it’s calculated

Online Archive storage is billed on GB-days (see Online Archive Billing):

Storage Cost (Annual) = (Archive GB) × (Rate $/GB-day) × (365)

Example calculation (10TB, US East / N. Virginia)

  • Archive GB = 10,240
  • Rate = 0.001578 $/GB-day
  • Annual cost = 10,240 × 0.001578 × 365 ≈ $5,898/year

(São Paulo uses a different $/GB-day rate, so the same formula yields a higher annual storage cost.)

2) Query Processing Cost (Annual USD)

What the number represents

Annual MongoDB charge for processing archived data during queries.
This is not the number of queries. It’s the annual $ cost derived from how much archived data is scanned/processed (see Online Archive Billing and Data Federation Billing).

Unit in table header

Query Processing Cost (Annual USD)

Why it’s intentionally not “queries per X”

MongoDB does not bill Online Archive by “number of queries.”
It bills by how much archived data is processed (scanned/read), because:

  • One query can scan 10 MB
  • Another query can scan 10 TB
  • Counting queries would be misleading

How it’s calculated

MongoDB charges:

  • $5 per TB processed
  • 10 MB minimum processed per query

For modeling, we estimate TB processed per month, then annualize:

Query Processing Cost (Annual) = (TB processed/month) × 12 × ($5/TB)

Usage levels used in the tables

  • Light: 2 TB processed/month
  • Moderate: 20 TB processed/month
  • Heavy: 200 TB processed/month

Example calculations

  • Light: 2 TB/month × 12 × $5 = $120/year
  • Moderate: 20 TB/month × 12 × $5 = $1,200/year
  • Heavy: 200 TB/month × 12 × $5 = $12,000/year

Cost factors (what drives TB processed)

  • Whether queries are bounded by the archive partition key (often a date; see Online Archive Overview)
  • Whether reports do broad “all-time” scans vs narrow time windows
  • Partitioning strategy and query patterns (tight filters reduce scanning)

3) Network Egress (Annual USD, Range)

What the number represents

Cloud provider network charge for data transferred out of the Atlas environment. This is not a MongoDB fee. It depends on:

  • Cloud provider (AWS/Azure/GCP)
  • Region
  • Destination (same-region vs cross-region vs internet/on-prem)
  • Transfer method and networking setup (peering/private endpoints vs public internet)

See: Atlas Data Transfer Costs.

Unit in table header

Network Egress (Annual USD, Range)

Why it’s a range

Without knowing the cloud provider + destination architecture, egress can’t be a single reliable number. The tables provide small/medium/large annual ranges as planning placeholders.

Egress levels used in the tables (illustrative)

  • Small: 100 GB/month exported/returned externally

  • Medium: 1 TB/month exported/returned externally

  • Large: 10 TB/month exported/returned externally

Annual egress ranges used in the tables (illustrative)

  • Small: $60–$180/year
  • Medium: $600–$1,800/year
  • Large: $6,000–$18,000/year

Important: If reporting and apps run in the same cloud and region (with private networking), egress may be minimal. Egress risk rises when data is routinely pulled to on-prem or exported in bulk.

Practical examples of when egress occurs

  1. On-prem BI (Tableau/Power BI server on-prem) querying Atlas and pulling large result sets → cloud-to-on-prem transfer.
  2. Bulk exports (monthly CSV/Parquet extracts from archived data) downloaded to laptops or internal file shares.
  3. Cross-region reporting (Atlas in US East; reporting app in São Paulo or Europe) → inter-region egress.
  4. Downstream data lake pipelines pulling archive results into another platform/account/region.

Annual Cost Estimates — US East (N. Virginia)

Storage rate: $0.001578 per GB-day
Storage-only baseline: ~$5,898 / year

Usage Level Storage Cost (Annual USD) Query Processing Cost (Annual USD) Network Egress (Annual USD, Range) Total Annual Cost Estimate (USD, Range)
Light + Small $5,898 $120 $60–$180 $6,078–$6,198
Light + Medium $5,898 $120 $600–$1,800 $6,618–$7,818
Light + Large $5,898 $120 $6,000–$18,000 $12,018–$24,018
Moderate + Small $5,898 $1,200 $60–$180 $7,158–$7,278
Moderate + Medium $5,898 $1,200 $600–$1,800 $7,698–$8,898
Moderate + Large $5,898 $1,200 $6,000–$18,000 $13,098–$25,098
Heavy + Small $5,898 $12,000 $60–$180 $17,958–$18,078
Heavy + Medium $5,898 $12,000 $600–$1,800 $18,498–$19,698
Heavy + Large $5,898 $12,000 $6,000–$18,000 $23,898–$35,898

Annual Cost Estimates — South America (São Paulo)

Storage rate: $0.002798 per GB-day
Storage-only baseline: ~$10,457 / year

Usage Level Storage Cost (Annual USD) Query Processing Cost (Annual USD) Network Egress (Annual USD, Range) Total Annual Cost Estimate (USD, Range)
Light + Small $10,457 $120 $60–$180 $10,637–$10,757
Light + Medium $10,457 $120 $600–$1,800 $11,177–$12,377
Light + Large $10,457 $120 $6,000–$18,000 $16,577–$28,577
Moderate + Small $10,457 $1,200 $60–$180 $11,717–$11,837
Moderate + Medium $10,457 $1,200 $600–$1,800 $12,257–$13,457
Moderate + Large $10,457 $1,200 $6,000–$18,000 $17,657–$29,657
Heavy + Small $10,457 $12,000 $60–$180 $22,517–$22,637
Heavy + Medium $10,457 $12,000 $600–$1,800 $23,057–$24,257
Heavy + Large $10,457 $12,000 $6,000–$18,000 $28,457–$40,457

When Network Egress Typically Applies (Examples)

Egress charges commonly occur when:

  • Reporting tools run on-prem and pull large result sets
  • Archived data is exported to files (CSV / Parquet) for downstream systems
  • Queries cross regions or clouds
  • Large historical datasets are downloaded by analysts or scripts

Egress is often minimal when:

  • Applications and BI tools run in the same cloud and region
  • Queries return aggregated metrics instead of raw rows

Summary & Key Takeaways

  • Archive storage is inexpensive and predictable
    • ~$5.9K/year (US East) or ~$10.5K/year (São Paulo) for 10 TB
  • Query Processing Rate ($5/TB processed) is the primary usage-based cost driver
  • Egress costs depend entirely on architecture and data movement patterns
  • Most teams land in Light–Moderate query ranges when queries are date-bounded and well-filtered
  • Broad historical scans and large exports are the main cost risk — and can be managed with design guardrails

This model provides a realistic planning baseline. With more detail on reporting frequency, query patterns, and where reporting tools run, the estimates can be further tightened.

Appendix Q&A

Does running reports against Online Archive impact the live Atlas cluster?

Short answer

Not in the way most people worry about.
Queries that read archived data are executed through Atlas Data Federation, not by consuming CPU, memory, or disk I/O on the Atlas cluster’s database nodes.

Reference:

What happens

  • Archived data lives in Online Archive (cloud object storage), not on the cluster’s local disks.
  • Queries that touch archived data are executed by the Data Federation query engine, which operates outside the cluster’s data-bearing nodes.

What can touch the cluster

If a query spans both hot (live) data and archived data:

  • The hot portion of the query runs on the Atlas cluster (just like any normal query).
  • The archived portion runs through Data Federation.
  • Results are merged transparently before being returned.

Practical implication:

  • Heavy historical reporting against Online Archive is designed to avoid starving transactional workloads on the live cluster.

“Archived-data reporting runs through Data Federation, so it doesn’t burn the live cluster’s CPU the way hot data scans do.”

How does data move from the live MongoDB cluster to Online Archive?

Short answer

You define the archive rule; Atlas moves the data automatically.

Reference:

How it works

  1. You configure an Online Archive on a collection and define archiving criteria (commonly date-based, such as createdAt or eventTime).
  2. Atlas continuously evaluates data in the background.
  3. Documents that meet the criteria are:
    • Copied into Online Archive
    • Removed from the hot tier
  4. The process is:
    • Online
    • Incremental
    • Fully managed by Atlas

Important characteristics:

  • No application changes required
  • No downtime
  • Archiving runs continuously, not as a one-time batch
  • Archived data is compressed and optimized for cost

“You define the rule; Atlas handles the movement.”

Can you run a query across both the live cluster and Online Archive?

Yes — this is a core capability.

Reference:

How it works

Online Archive is exposed as a logical extension of your collections. A single query can span:

  • Hot data in the Atlas cluster
  • Archived data in Online Archive

MongoDB automatically routes:

  • Hot reads to the cluster
  • Archive reads to Data Federation

Conceptual example

db.events.find({
  eventTime: { $gte: ISODate("2022-01-01") }
})

If recent data is hot and older data is archived:

  • Recent documents are read from the cluster
  • Older documents are read from Online Archive
  • Results are merged transparently

Why this matters

  • No separate query language
  • No ETL just to query history
  • Existing reports can continue to work without modification

“It’s one query, one namespace — MongoDB handles where the data lives.”

Practical considerations

  • Archived data is optimized for throughput and cost, not ultra-low-latency transactional access.
  • Query Processing cost is driven by how much archived data is scanned, not by query count.
  • Best practices:
    • Always filter on the archive partition key (usually a date)
    • Avoid unbounded “all-time” scans
    • Aggregate early where possible

Reference links

  1. Atlas Online Archive Billing (includes $5/TB processed + 10MB minimum; notes transfer considerations)
    https://www.mongodb.com/docs/atlas/billing/online-archive/
  2. Atlas Data Federation Billing (explains “data processed” billing at $5/TB and how scanning drives cost)
    https://www.mongodb.com/docs/atlas/billing/data-federation/
  3. Atlas Data Transfer Costs (transfer/egress varies by provider/region/source/destination; invoices are source of truth)
    https://www.mongodb.com/docs/atlas/billing/data-transfer-costs/
  4. Online Archive Overview (what Online Archive is and how it works conceptually)
    https://www.mongodb.com/docs/atlas/online-archive/overview/
  5. MongoDB Pricing Page (entry point for Online Archive pricing tables by region)
    https://www.mongodb.com/pricing
  6. Atlas Data Federation (overview of the feature underpinning archive querying)
    https://www.mongodb.com/docs/atlas/data-federation/