Skip to main content

Serverless to Dedicated Costs Explained

Mary Gorman avatar
Written by Mary Gorman
Updated over 2 weeks ago

Description

Creation of Atlas serverless instances was deprecated in late Jan 2025. Existing Serverless clusters would be migrated to an Atlas dedicated cluster if:

  • The Serverless Instance data size exceeded 5GB prior to the migration date

  • The Serverless Instance project had at least one Serverless private endpoint enabled

Serverless instances with less than 5GB data size and no project serverless private endpoints were migrated to flex tier clusters or free tier (M0) clusters.

Migration

Atlas dedicated clusters were deployed in the same Atlas project and region as the serverless instance, with sufficient CPU and IOPs for the migration and auto-scaling enabled following the migration (so that the cluster can downsize to the appropriate size as determined by our auto-scaling algorithm).

Dedicated clusters with pre-determined tiers, storage capacities and auto-scaling ranges were deployed for the duration of the migration depending on the migrated data size. Post-migration the dedicated cluster was allowed to auto-scale down in response to CPU and memory usage over the subsequent days.

The below table shows the chosen dedicated migration clusters tier(s) together with the storage capacities, auto-scaling ranges, auto-scaling duration and expected (ending) default cluster tiers:

Migration Cluster

Storage Capacity

Auto-scaling Range

Auto-scale Duration

Default Cluster Tier

M30

40GB

M10 to M30

2 days

M10 with 40GB

M40

80GB

M10 to M40

3 days

M10 with 80GB

M50

160GB

M20 to M50

3 days

M20 with 160GB

M80

750GB

M40 to M80

3 days

M40 with 750GB

Note: The duration of the migration depended on the data size migrated and the CPU and IOPs of the destination cluster deployed, however, most migrations completed in less than 60 minutes with some migrations requiring several hours to complete.

Post Migration

Following the serverless to dedicated cluster migration:

  • The Serverless Instance is terminated

  • The Serverless Instance backups are deleted

The new Atlas dedicated cluster was:

  • Deployed with the same name as the original serverless instance

  • Deployed with MongoDB version 8.0 (same as that of the Serverless cluster)

  • Deployed on the same cloud provider and region as the original serverless instance

  • Deployed with tier auto-scaling and storage auto-scaling enabled by default

  • Deployed with cloud backups enabled by default (without continuous cloud backups)

  • Auto-scaled down to the optimum cluster tier size based on CPU and memory usage

Following the serverless to dedicated migration the new dedicated cluster retained the storage capacity configured for the migration cluster. Ie. If an M30 with 40GB was deployed as the migration cluster, post migration the default cluster would be an M10 with 40GB storage capacity.

Migration Costs Explained

Costs are incurred from the serverless instance to dedicated migration(s):

  • Migration dedicated cluster with adequate CPU and IOPs for migration

  • Interim dedicated cluster(s) deployed during tier auto-scaling

  • Data Transfer Costs

Example:

A serverless instance deployed on the AWS cloud provider in US_EAST_1 (ie. North Virginia) region with a data size 50GB is expected to incur a migration cost circa $95 as per the following cost breakdown:

Cost

Source

Price

Cost

Migration Cluster

M50 cluster with 160GB

$2.00 x 24 hours

$48.00

Interim Cluster 1

M40 cluster with 160GB

$1.13 x 24 hours

$27.12

Interim Cluster 2

M30 cluster with 160GB

$0.63 x 24 hours

$15.12

Maximum Data Transfer

50GB

$0.01 x 50GB

$5.00

Migration Cost

$95.24

Following the migration the dedicated cluster is auto-scaled down to the new M20 with 160GB storage capacity on AWS cloud provider in the US_EAST_1 (ie. North Virginia) region after three days. Unless modified the new M20 dedicated cluster will continue to have an hourly rate of $0.29 per hour thereafter.

Note: As the default storage capacity of an AWS M20 instance size is US_EAST_1 (ie. North Virginia) region is 20GB, costs incurred for storage above the default, although included in the hourly rate, are shown separately in Atlas bills. See also Atlas Storage Capacity billing documentation.

Reducing Post Migration Costs

To reduce costs post migration:

Reduce the storage capacity. If the configured storage capacity of the new dedicated cluster is greater than the “Disk Space Used” size in the cluster’s metrics, modify the cluster to scale down the storage capacity.

Modify the backup policy. Newly converted clusters are configured with cloud backups and the default backups policy. Modify the backup policy to retain less backups than the estimated default number snapshots.

Disable Cloud Backups. If retained backups are not required modify the cluster to disable configured cloud backups.

Adjust the tier auto-scaling range. Following the serverless to dedicated migration the dedicated cluster will scale down to the lowest instance size in response to CPU and memory usage in the days following the migration. If your expected CPU and memory usage is typically below the lower auto-scaling range, modify the cluster to allow the cluster to scale below this instance size.

Migrate to Flex tier or free tier (M0) or Flex cluster. A dedicated cluster can be migrated to a flex tier cluster or free tier (M0) instance size, using the MongoDB database tools mongodump and mongorestore, if the storage capacity of the cluster is below the following thresholds:

  • Less than 5GB for Flex tier

  • Less than 512MB for free tier (M0) instance size

Terminating the dedicated cluster. If the dedicated cluster is no longer required, terminating the cluster will cease future incurred costs.

Did this answer your question?