Evan Daniel wrote: > If we're knowingly misrouting around slow nodes, then it seems to me > we should make a specfic effort to have the one request that can go to > the slow node be the one that it is most likely to be able to serve, > instead of the one that happens to arrive first.
I agree in principle, but the question is how to do that without introducing long delays - how long should you wait for a request that's closer to the slow peer's location? Could we give slower peers a smaller region of the keyspace, not by modifying the swapping algorithm but by modifying the algorithm that determines which peer is closest to a given key when routing a message? For example, instead of circular distance could we use circular distance over speed? Again, there would probably have to be limits to prevent a fast peer taking over the whole keyspace... Cheers, Michael
