What we want to test

[victorb]

There is bunch of good things we should test from these two issues:

• Leveling up testing infrastructure ipfs/notes#202
• go-ipfs biweekly releases ipfs/notes#203

It would be nice to have a concise list of things we want to be able to test.

Things like:

• Adding a file on one node, cat'ing it on 100, measuring how long time it takes to finish
• Measure and compare result of ^ across versions and implementations (go-ipfs vs js-ipfs)
• How much memory we end up taking if we add/rm then repo gc on nodes
• How many file descriptors we end up with if we just stay idle with a node

And so on

[victorb]

Also, we should figure out what we need to expose on the debug/metrics/prometheus endpoint in the ipfs implementations to support this.