Is it usually the case that we pair pessimistic detection with lazy versioning and optimistic with eager?
paluri
@russt17 I think this is backwards, see slide 49. Although, looking at slide 51, it seems like you really can do whatever makes you happy.
ekr
Intuitively, it seems like pessimistic detection would be better paired with eager versioning. We saw a few slides ago that eager versioning is more efficient, but less conflict-tolerant; if the detection is pessimistic, we won't have to worry about conflicts because we check before every write.
Similarly, if we're using optimistic conflict detection, we're probably in a context where there won't be many conflicts. In this case, it makes sense to use lazy versioning, because it's not very likely that we'll have to wipe the buffer.
Is it usually the case that we pair pessimistic detection with lazy versioning and optimistic with eager?
@russt17 I think this is backwards, see slide 49. Although, looking at slide 51, it seems like you really can do whatever makes you happy.
Intuitively, it seems like pessimistic detection would be better paired with eager versioning. We saw a few slides ago that eager versioning is more efficient, but less conflict-tolerant; if the detection is pessimistic, we won't have to worry about conflicts because we check before every write.
Similarly, if we're using optimistic conflict detection, we're probably in a context where there won't be many conflicts. In this case, it makes sense to use lazy versioning, because it's not very likely that we'll have to wipe the buffer.