examples: add a benchmark end-to-end example

[tzach]

Such an example should:

• start a scylla cluster with private IPs
• start loader instances with cassandra-stress
• execute VPC peering between the loaders and ccylla cluster
• start the loaders

For a similar example in Ansible see
https://github.com/scylladb/cassandra-test-and-deploy

[tzach]

Example for Alternator with loaders from @nuivall
https://github.com/nuivall/alternator-dynamo-bench

[rjeczalik]

@nuivall I need your opinion on this, what do you think is best way to approach this?

I have created some time ago Terraform modules for benchmarking - terraform-aws-scylla-bench (complementary to a AWS deployment module - terraform-aws-scylla).

What are your intentions with alternator-dynamo-bench, do you want to unify those efforts somehow, or you would like to keep them separate?

There are a lot of pros/cons regarding both directions, thus the question.

[nuivall]

The goal for alternator-dynamo-bench was just to present some alternative. Currently I am not planning to improve it, unless it is needed for something (one example could be if we publish alternator results and want to bundle them with some easily reproducible solution, but it won't happen in a near future).

Long term I think a modular solution which allows to choose which benchmark and which deployment (cloud vs direct aws) could be useful.

[tzach]

lets start with the popular cassandra-stress for the loader.
We can change it after

[rjeczalik]

sent a draft - #57