You are on page 1of 1

Lecture 2 4-1-20 Questions

1) Does MapReduce have to create extra copies of data or move data around in order
to group data by key after mapping?

2) I don't see how synchronizing with replicated state machines solves the issue of
updates to replicated masters. If the state machines need to get updates in the
exact same order to reach the same final state, then it seems like the fundamental
issue of replicated masters seeing events in different orders is still there right?

- There are two necessary things in the solution: Make sure that updates
happen in the same order, and make sure that given a current state and an update,
the next state to go to is deterministic

3) For ordering updates based on absolute timestamp, how can you know that an
update chronologically before your latest update is just being delayed by the
network?

- This can be solved by logical timestamps. Absolute timestamps cannot solve


this issue.

You might also like