Ask Your Question
3

AWS Resource Allocation Limits for PAT Analysis

Does anyone know what are the resource requirements to deploy OpenStudio-server on AWS and run a PAT simulation in cloud server mode?

I'm having trouble with changing the AWS resource types other than what's recommended by NREL. I can successfully run a simulation using the settings stated in the GitHub page. The main failure occurs when trying assigning pods after changing the AWS instance type. For example, I tried changing t2.xlarge to t4g.xlarge instance which seems to be similar.image description

I would like to know what instance types we could use and what are the minimum requirements to consider for the following resources when allocating them.

  1. Minimum number of CPU's in a Node
  2. Minimum number of Nodes
  3. Minimum Storage for a Node (Default is 80GB)
  4. openstudio cluster: Minimums of nfs-server-provisioner.persistence.size and db.persistence.size (Default is 200GB)
Kushan_26's avatar
71
Kushan_26
asked 2021-11-23 20:47:10 -0500
__AmirRoth__'s avatar
4.4k
__AmirRoth__
updated 2022-02-23 12:20:14 -0500
edit flag offensive 0 remove flag close merge delete

Comments

add a comment see more comments

1 Answer

6

Hi and thanks for the questions.

1 ) Currently, the minimum number of nodes is 3x t2.xlarge instances or equivalent sized node (4 vcpu, 16gb mem per node). We are working to slim this down a bit to get everything to fit on 2 t2.xlarge instances.

2 ) 3 t2.xlarge or equivalent.

3 ) Ephemeral storage per node could prob be less than 80GB but I would stick with this.

4 ) The defaults should work fine, but if you are running a very large analysis, or this is a long running cluster you can increase these depending on how much data is being written to the volume.

As for your question about switching to t4g.xlarge, I don't see that as problem as far as helm/k8s is concerned, but I would make sure this is node instance type is compatible with EKS in AWS. If you post the AWS error I'd be happy tp help where I can. I am assuming that this is AWS related problem and not with the server-helm chart.

tijcolem's avatar
161
tijcolem
answered 2021-11-30 14:57:13 -0500
edit flag offensive 0 remove flag delete link
more

Comments

Hello, Thank you for verifying the resource limits. I did some research about t4g instances and looks like they are not supported by AWS stacks. AWS is working on adding support for these instances. [https://repost.aws/questions/QUZX2cCw...]

Kushan_26's avatar Kushan_26 (2021-12-07 12:15:39 -0500) edit
more
  • delete
add a comment see more comments