Burstable M10/M20 tiers
Parent: Atlas Cluster Topology and Tier Economics · Published reference · snapshot 2026-10-20 · skill ranked-5130ee0b
↓ Facts as markdownall context files
Source-anchored research for Burstable M10/M20 tiers. 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's System: CPU (Steal) alert is scoped to AWS EC2 burstable M10/M20 instances and applies when the instance credit balance is exhausted; it measures CPU usage above the guaranteed baseline credit-accumulation rate. It should not be treated as a provider-neutral M10/M20 signal. (high confidence) (verified as of 2026-10-20) [source]
- AWS EC2 documents two burst-credit modes: Standard instances return to baseline when accrued credits run out, while Unlimited instances can continue bursting on surplus credits and incur additional charges when average CPU over the rolling 24-hour period exceeds baseline. (low confidence) (verified as of 2026-10-20) [source]
Measurements and reference values
- Atlas bases M10/M20 auto-scaling CPU decisions on Relative System CPU Utilization, calculated by dividing Normalized System CPU by the provider baseline. MongoDB says the baseline is typically 20%-50% for burstable types and is not shown in Atlas metrics; provider-specific auto-scaling criteria are subject to change. (high confidence) (verified as of 2026-10-20) [source]
- In a load-test discussion dated October 21, 2022, an Atlas user reported that a test spent most of its time on M10/M20, saw CPU Steal around 130%-140% with poor performance, and became smoother after running long tests on M30. This is one workload report, not a general performance benchmark. (low confidence) [source]
- A MongoDB feedback item for M10/M20 CPU-credit metrics and alerts was still marked Submitted when checked on October 20, 2026. A commenter dated July 25, 2025 reports CPU Steal rising from 0% to 80% within minutes; this records a user request and report, not confirmation that the requested metrics are absent from every Atlas interface. (low confidence) (verified as of 2026-10-20) [source]
Problems, failure modes and limitations
- Atlas says M10 and M20 use burstable performance infrastructure and warns that cloud providers can cap CPU after burst periods. AWS and Azure describe the credit model: instances earn credits below baseline, spend them above baseline, and run at baseline when credits are exhausted. This makes sustained above-baseline demand a throttling risk. (high confidence) (verified as of 2026-10-20) [source]
Comparisons and alternatives
- Inference from Atlas's low-traffic positioning and the provider credit models: M10/M20 fit intermittent CPU peaks when sustained usage stays near or below baseline; a continuously CPU-heavy workload can exhaust burst headroom and trade the lower tier for baseline-limited throughput. Compare measured workload behavior with a higher tier rather than treating a short burst test as steady-state capacity evidence. (medium confidence) (verified as of 2026-10-20) [source]
- Microsoft documents that Azure B-series VMs accumulate CPU credits below base performance, spend them above base, and throttle back to base performance when credits are exhausted. Atlas documents an Azure-specific M10/M20 auto-scaling signal but does not identify B-series as the Atlas host family, so the B-series behavior is provider context rather than a confirmed Atlas host mapping. (medium confidence) (verified as of 2026-10-20) [source]
Facts and statements
- Atlas positions M10 and M20 as dedicated tiers for development and low-traffic production; they support replica sets only, while Atlas recommends M30+ for high-traffic production. (high confidence) (verified as of 2026-10-20) [source]
- A practitioner article edited January 17, 2025 reports that Atlas M10/M20 on Azure used B-series VMs and that the Atlas CPU Steal metric was unavailable there at the time. Atlas's current alert documentation separately scopes its CPU Steal alert to AWS EC2; treat the Azure VM mapping as a dated user report, not a current Atlas tier-to-VM guarantee. (medium confidence) (verified as of 2026-10-20) [source]
- Atlas documents M10 and M20 for development and low-traffic production, limits them to replica-set deployments, and recommends M30 or higher for production; M30+ also supports sharded deployments. (high confidence) (verified as of 2026-10-20) [source]
- Atlas says M10/M20 use burstable performance infrastructure. Cloud providers can cap CPU after burst periods, which can throttle heavy workloads; MongoDB also warns that sustained M10/M20 loads may degrade performance. (high confidence) (verified as of 2026-10-20) [source]
Children
- No children recorded.