- Newest
- Most votes
- Most comments
Write throughput is not relevant. gp2 IOPS accounting is a function only of the IOPS used, not throughput.
gp2 provides 5.4 million burst i/o credits at 3000 IOPS.
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/EBSVolumeTypes.html
When you use up this burst you will get throttled to baseline:
https://aws.amazon.com/blogs/database/understanding-burst-vs-baseline-performance-with-amazon-rds-and-gp2/
Your workload is clearly write heavy. The reason you see so few read IOPS is likely that many of these SELECTS are being served from the buffer cache and not disk. Your instance has 4 striped EBS volumes so while you have a very high number if burst iops, your workload is continually burning through this burst balance and then you get throttled to 1500 IOPS.
To avoid getting throttled you could consider increasing the size of your allocated storage, since gp2 provides 3 IOPS per gb, you can do the math to allocate a volume to meet your minimum IOPS requirements. Or you could consider io1 and specify the precise IOPS you require
-Phil
Relevant content
- Accepted Answerasked a year ago
- What happens to Amazon RDS and Amazon Redshift queries that are running during a maintenance window?AWS OFFICIALUpdated 9 months ago
- AWS OFFICIALUpdated 6 months ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 2 years ago