> I suspect (but cannot prove) that performance would jump for systems
> like ours if WAL was done away with entirely and the individual data
> files were synchronized on commit.
You know.. thats exactly what WAL is designed to prevent? Grab a copy of
7.0 and 7.1. Do a benchmark between the 2 with fsync on in both cases.
I think you'll find one is about 50% faster than the other on a single
disk system, and about the same if you have gobs of battery backed write
cache.