Autoscaling using AWS t3a EC2 Instances with Unlimited Mode

I noticed that t3a Instances are being used in default AWS Config https://github.com/OpenDroneMap/ClusterODM/blob/master/docs/aws.md.

I noticed these instances are launched in Unlimited Mode by default. Has anyone done an analysis on cost for running them in Standard Mode vs Unlimited Mode. AWS docs suggest t3a instances being used at high CPU utilization for short periods of time should be launched in Standard Mode.
image

Graph of a sample 21 minute job run on a t3a.small instance:
image
image

2 Likes

Seems like docker-machine does not support specifying instance credit specification. https://docs.docker.com/machine/drivers/aws/

Thankfully, AWS UI has an option to set default credit specification for new instances.
image

3 Likes

Awesome. Thanks for sharing!

I’ve done some calculations and it seems that using m5 instances is a better deal than using t3a instances wherever possible. Let me know if you want to update it in the repo. Would be happy to send a PR.

PR away! Thanks!

Sent! https://github.com/OpenDroneMap/ClusterODM/pull/61

2 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.