Benchmark state machine logic

[adrianbrink]

Having released new staking to a public testnet we noticed that it slows down the block production. We should have a framework that allows us to benchmark the per block actions. This will allow us to quickly identify performance issues and optimize reads/writes.

@rigelrozanski

[cwgoes]

One option: pipe statistics to Riemann, which has an official Go client, for easy visualization and debugging.

[cwgoes]

I think we can use the same benchmark for this and for gas (#1008).

[jackzampolin]

My friends over at InfluxDB wrote this tool to debug their own datastore: https://github.com/influxdata/grade I would suggest this! But I think regardless this is an important piece of hardening.

[jackzampolin]

Closing this issue, will track testing improvements elsewhere.

[rigelrozanski]

Where?

[alexanderbez]

We don't have any direct state-machine benchmarks. Our closest bet would be simulation.

[cwgoes]

I think this is absolutely something the SDK should include. When we want to scale e.g. Gaia to high TPS we'll need to spend a lot of time optimizing transaction execution.

[elias-orijtech]

Similar to or duplicate of #13981?

[tac0turtle]

they seem different one is add telemetry and this one is to do bechmarks, want to work on this @elias-orijtech