Exoscale cloud provider support?

[Zhatra]

Hi! I was wondering if there's any support for Exoscale in PerfKitBenchmarker? I noticed there's support for AWS, Azure, GCP, and even CloudStack, but I couldn't find anything for Exoscale.

This seems like it could be a relatively straightforward addition since:

PKB already has CloudStack support, and Exoscale was historically CloudStack-based (though they've moved to their own API v2)
Exoscale's driver in Apache Libcloud is actually based on the CloudStack driver
They offer S3-compatible storage and compute instances similar to other supported providers

If there isn't support yet, could someone point me in the right direction for implementing it? I'd be interested in contributing this if it's something the project would accept.

Thanks!

[s-deitz]

Thanks for the inquiry. We do take contributions to add other providers but often are unable to test the code.

Note that contributions do require signing a Contributor License Agreement, so make sure you are okay with that.

I would then recommend looking at a cloud provider that only implements the basics such as Alicloud. After that, it is mostly a matter of implementing a set of classes to represent fundamental resources like virtual machines, disks, and a network. The simplest benchmark is called cluster_boot and is easy to run. On GCP, this may look like:

./pkb.py --cloud=GCP --benchmarks=cluster_boot --machine_type=e2-micro --zone=us-central1-a

Note that especially when developing for a new cloud, you should make sure that you tear down the resources since they will accrue costs if PKB fails to tear down the resource or that part isn't implemented yet.

When you have cluster_boot up and running, I think that would be a good time to get the code committed. We can review smaller PRs more quickly.

Let us know if you are going to take this on and if you have further questions.