On Thu, 15 Apr 1999, Luke Mewburn wrote:
It now takes 2-3 seconds instead of 2 minutes, because I changed from using usernames to using UID's.
Interesting. You just plunked in the UIDs instead of usernames with no other modifications, right?
I only have a default quota and it still takes a while to do a quota on. I wonder how the mechanism for default quotas looks in contrast to a "named" quota. The box doesn't even have a password table so the quota on should be just time spent to scan the filesystem.
Now that I think about it, what would be cool is to use the functionalty of WAFL to speed up quota updates. All that would have to be done is update quotas with metadata that changed from the last snapshot when the quota was on. Of course one would have to write the quota database to disk, at least on shutdown. This could be very fast, unless one needs to traverse the filesystem in search of the hierarchical placement of those inodes.
This brings an interesting point, how are quotas counted for hard links? What about hard links across qtrees. This should be possible. Are they counted in each qtree? That would be a sensible thing to do, but it is certainly misleading.
Tom