Opportunity Name:
EBS Volume Cost Optimizations (io1, io2 to gp3)
AWS Resource Type:
EBS Volumes
Opportunity Description:
This opportunity retypes io1/io2 volumes to cheaper GP3 volumes with equivalent performance.
Criteria for identifying the opportunity:
The Finder identifies EBS volumes that meet the following criteria:
- Current volume type is io1 or io2
- Provisioned IOPS is 16,000 or less
- Volume has been in use for at least 7 days
- Volume is attached to a single instance (multi-attach volumes are excluded)
- io2 Block Express volumes are excluded
Potential savings (range in % on annual basis):
GP3 volumes are 36% cheaper than io1/io2 volumes in storage size, include 3,000 provisioned IOPS, and are up to 13x cheaper in terms of additional provisioned IOPS. Although GP3 volumes are charged separately for storage throughput, the storage size and provisioned IOPS cost savings are much more significant. As an example, a 1TB volume in the US East 1 region with 10,000 provisioned IOPS and 1000MiBps throughput costs $1,590 per annum for GP3 vs $9,300 per annum for io1/io2 (87% savings)
What happens when the Fixer is executed?
The Fixer first takes a snapshot of the volume and then retypes the volume from io1/io2 to GP3 with equivalent provisioned IOPS and throughput settings. Once the retype operation has been completed successfully, the snapshot is deleted.
IOPS and throughput for the new gp3 volume are not simply set to a default value. CloudFix analyzes the volume’s actual historical peak usage via CloudWatch metrics over a configurable lookback period and adds a safety buffer, then provisions gp3 to match or exceed that observed peak. If usage metrics aren’t available, it falls back to matching the volume’s current io1/io2 IOPS and throughput settings. Either way, the gp3 volume is provisioned to perform at least as well as the original.
Is it possible to rollback once CloudFix implements the fixer?
Yes - it is possible to manually retype the volume back to the original io1/io2 type without service disruption. In the rare event of a failure during the retype procedure, the volume can be manually restored from the snapshot taken by the Fixer.
Can CloudFix implement the fix automatically once I accept the recommendation?
Yes
Does this fix require downtime?
No - any EBS volume with the Elastic Volumes feature supports volume type modification without detaching the volume. All volumes attached to an instance after November 3, 2016, 23:40 UTC support Elastic Volumes.
The application does not need to be paused or otherwise prepared for interruption: the volume type change happens online, without restarting the instance or detaching the volume, so any production application depending on it can keep running throughout the migration.
Priyanka Bhotika
Comments