Load Tests - Etch storage

[mikera]

Things to test:

• Large DB size
• Performance of caching in large DBs
• Performance during GC
• Performance under concurrent access

[helins]

I'd like to tackle this sooner than later from two front:

Integrity. Since Etch is append-only, I really don't see why behavior would become dysfunctional with a higher volume of data. Still, now that I've implemented extensive cell generators (using test.check), I'd like to prove empirically that things just work even with ~1TB of "high-entropy" data.

Actual performance. Focus on actual R/W speed as a function of max JVM memory and DB size, minimizing the overhead of generating cells.

Sounds simple but both present a set of challenges so this won't be exactly trivial to solve.