It's nice but I think you need to test on more than 2 instances per provider. Things can vary _a lot_ and I think benchmarks like this should collect a lot more

It's nice but I think you need to test on more than 2 instances per provider. Things can vary _a lot_ and I think benchmarks like this should collect a lot more data. With Ansible and such it's not too difficult to set up and run a benchmark of 100 instances per provider and it's not overly expensive either. I've even seen it recommended that when starting an instance on AWS, you spawn 100 instead and take the one that performs best.