Atlas offers two database editions. Both editions run MongoDB and use the same interface, drivers, APIs, and most operational workflows. They differ in how they provision compute and storage:
Atlas Core: the Atlas cluster architecture where compute and storage run on the same node. It is available across Free clusters, Flex clusters, and dedicated cluster tiers.
Atlas Infinite: an architecture that separates compute from storage into independent layers, designed for mission-critical high-performance use cases. Atlas Infinite is available as a public preview, and both the product and its documentation might change during the preview period.
The following diagram compares the two editions.
Atlas Core couples compute and storage on each node, while Atlas Infinite runs compute and storage as separate layers
Compare Editions
Aspect | Atlas Core | Atlas Infinite |
|---|---|---|
Architecture | Compute and storage run on the same node. Each data-bearing node runs on a virtual machine with attached storage. | Compute and storage run as independent layers. MongoDB manages the storage layer separately from the compute nodes. |
Compute scaling | You change compute capacity by selecting a different cluster tier. Atlas restarts the cluster's nodes when you change the tier, with no downtime on dedicated clusters. | You scale the cluster tier independently of storage. Scaling doesn't move your data, so it is fast. |
Storage scaling | You scale storage capacity and Input/Output Operations Per Second (IOPS) independently of compute on dedicated clusters, up to the maximum for the cluster tier. To add capacity beyond that maximum, you scale to a higher tier. | Storage capacity increases automatically based on usage, independently of the cluster tier, without downtime. |
Horizontal scaling | You shard a cluster when its data approaches the storage capacity, and add workload-specific nodes, such as read-only or analytics nodes, as needed. | Each highly available shard (HA shard), a unit distinct from a MongoDB shard, stores up to 128 TB of logical data without requiring sharding. Adding workload-specific nodes is much faster. |
Recovery | Atlas monitors your database regularly, and performs auto-healing if it discovers an issue. If it replaces a node, the new node performs an initial sync before it serves queries. | Atlas monitors your database regularly, and performs auto-healing if it discovers an issue. If it replaces a node, the new node immediately connects to the storage layer, with no need for an initial sync. |
Cloud provider and tier availability | Free clusters, Flex clusters, and dedicated cluster tiers, on AWS, Azure, and Google Cloud. | Public preview: |
Feature coverage | Supports the full Atlas feature set. | Some Atlas features aren't supported during public preview. |
Choose an Edition
Consider Atlas Core when:
You want the full Atlas feature set
You need multi-region or multi-cloud clusters
You need sharded clusters
Consider Atlas Infinite when:
You want to scale compute and storage independently
Your workload has spikes that are difficult to preprovision for
You want to grow storage capacity without scaling the cluster tier
You want faster failover and recovery
Confirm that your workload fits the public preview limitations first.
The Atlas Infinite architecture describes how the compute and storage layers interact.