Using VRFs to randomly assign stakers to shards

Sharding
2

Nice writeup Paul! I would just add that if a staker wishes to withdraw, they should be required to remain for at least one full “shift” following their request. Otherwise, a staker who is targeting a particular shard could do “better than random” by withdrawing whenever they’re about to be assigned to a different shard.

I think of VRF-based randomness as being very similar to RANDAO, except that determinism removes the need for commitments. So if we want to reduce biasability, we can use similar techniques:

  1. Batch k VRFs together, and use the result for the next k blocks
  2. The subcommittee scheme you proposed
  3. Avalanche RANDAO

We’re planning to use #1 for our project. Its biasability is worse than #2 and #3, but it’s still very low if we use a large batch size, and we prefer to avoid public key encryption.

1 Like