M10 shared-vCPU dedicated tier economics
Parent: Atlas Flex Pricing and Flex-vs-Dedicated Break-Even · Published reference · snapshot 2026-10-23 · skill mongodb-cohort-b7c969c2a118
↓ Facts as markdownall context files
Source-anchored research for M10 shared-vCPU dedicated tier economics. 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
- Atlas calls M10/M20 Dedicated because each mongod runs on its own instance; this does not promise exclusive physical cores or sustained full vCPU capacity. The pricing FAQ explicitly calls their vCPUs shared, and both official pages warn about burstable infrastructure. (high confidence) (verified as of 2026-09-29) [source]
Measurements and reference values
- The September 29, 2026 pricing table lists AWS us-east-1 M10 at $0.08/hour with 2 GB RAM, 2 listed vCPUs and 10 GB default storage. At that rounded rate, 730 hours is $58.40; the page separately advertises $56.94/month. Neither figure is a universal all-in quote. (low confidence) (verified as of 2026-09-29) [source]
- For M10/M20 diagnosis, Atlas uses Relative System CPU Utilization: Normalized System CPU divided by provider baseline CPU utilization. The baseline is not exposed in Atlas metrics and is typically 20%-50% for burstable instances. A modest-looking CPU percentage can therefore accompany an auto-scale decision. (medium confidence) (verified as of 2026-10-19) [source]
- Self-hosted comparisons should include operations work as well as infrastructure. C3X enumerates upgrades, patching, restore verification and failover testing as costs that can absorb apparent managed-service savings. Its labor-hour and break-even estimates are scenario assumptions, not measured universal thresholds. (low confidence) [source]
- The pricing page dated September 29, 2026 lists M10 at $0.08/hour in its reference regions: AWS us-east-1 has 2 GB RAM and 2 vCPUs; Azure eastus2 has 2 GB and 1 vCPU; Google us-central1 has 1.7 GB and 0.5 vCPU. These are provider-specific base configurations, not equivalent sustained compute guarantees. (low confidence) (verified as of 2026-09-29) [source]
- Google documents shared-core VMs as time-sharing physical CPU and allowing temporary bursts; for example, g1-small sustains up to 50% of one vCPU. This independently establishes the fractional-capacity mechanism, but does not prove that every current Atlas Google M10 uses g1-small. (low confidence) (verified as of 2026-10-21) [source]
Problems, failure modes and limitations
- A performance-cost investigation should compare representative sustained workloads and correlate slow-operation periods with available CPU and throttling signals. This is a synthesis from Atlas's burst warning, baseline-based scaling model, and a practitioner upgrade report; a short burst test cannot establish sustained M10 throughput per dollar. (medium confidence) [source]
Comparisons and alternatives
- MongoDB positions M10/M20 for development and low-traffic applications; its pricing FAQ narrows them to small, non-critical workloads. M30+ is the recommended starting point for performance-critical, high-traffic production. These statements scope suitability rather than prohibit all M10 production use. (high confidence) (verified as of 2026-09-29) [source]
- Using the same AWS table, M30 costs $0.54/hour versus M10's $0.08/hour: 6.75 times the base rate, or $394.20 versus $58.40 over 730 hours. This arithmetic compares tier prices, not measured throughput per dollar or identical CPU guarantees. (medium confidence) (verified as of 2026-09-29) [source]
- Default storage is included in the running cluster's base charge; extra storage configuration, backup and transfer can change the bill. Calling the M10 sticker price 'compute alone' is therefore incomplete. Price a specific deployment rather than adding the entire default storage charge again. (high confidence) (verified as of 2026-10-20) [source]
- Entry-tier base pricing is not a total-cost ceiling: customized storage changes pricing, and enabled cloud backups incur separate storage and restore-transfer costs. Compare the configured deployment rather than a bare advertised monthly entry price. (high confidence) (verified as of 2026-10-20) [source]
- Methodological inference: assess entry-tier economics with sustained workload tests and time-aligned CPU and scaling evidence, rather than inferring steady-state throughput from advertised vCPU count or a short burst. Atlas warns that sudden workload spikes may finish before auto-scaling provisions capacity. (medium confidence) (verified as of 2026-10-21) [source]
Facts and statements
- AWS burstable instances accrue credits below baseline and spend them above it. Standard mode returns toward baseline after credits run out; Unlimited mode can continue using surplus credits with possible additional charges. AWS defaults do not establish which mode or underlying instance Atlas uses. (low confidence) (verified as of 2026-10-20) [source]
- In a February 24, 2025 Meteor forum post, one practitioner reported resolving development Atlas Search performance problems by upgrading to M30. They quoted support identifying CPU throttling and non-zero CPU Steal on M10/M20. This is one workload report, not a tier-wide throughput benchmark. (low confidence) [source]
- Deyan Petrov's engineering account, edited January 17, 2025, reports Azure M10/M20 using burstable B-series VMs and lacking the CPU Steal visibility the author wanted. Treat the VM mapping and observability gap as dated practitioner observations, not current Atlas guarantees. (low confidence) (verified as of 2025-01-17) [source]
Children
- No children recorded.