Impact of Adversarial Transactions on Scheduling Algorithm

We need to understand what an adversary can do by injecting worst-case transactions on the proposed scheduling algorithm (https://github.com/oasislabs/ekiden/pull/749).

The plan here is to write a tool to modify the synthetically generated load to inject what we think worst-case transactions might be wrt throughput. Once we have this, we can run the generated transactions through the scheduling algorithm and see what the impact is on the overall throughput.

Details

The scheduling algorithm is described here: https://github.com/oasislabs/rfcs/pull/25

A risk here is that we don't really know what a worst-case transaction looks like for any particular scheduling algorithm, so we might end up showing that the scheduling algorithm is okay wrt some bad inputs, but not the actual worst-case ones.

Acceptance Criteria

This is experimental in nature. This is done if it shows that the scheduling algorithm is resistant to injecting bad transactions, or if it has problems and needs to be improved.