Subject: RE: namei caching of newly created files?
To: der Mouse <mouse@Rodents.Montreal.QC.CA>
From: Gordon Waidhofer <email@example.com>
Date: 01/19/2005 23:33:30
> -----Original Message-----
> From: tech-kern-owner@NetBSD.org [mailto:tech-kern-owner@NetBSD.org]On
> Behalf Of der Mouse
> Sent: Wednesday, January 19, 2005 10:23 PM
> To: tech-kern@NetBSD.org
> Subject: Re: namei caching of newly created files?
> >>> IMO, once NFS rules are conceded, might was well go hog wild with
> >>> write-back caching. Why not. Loss is loss.
> >> But there's loss and then there's loss. [...]
> > Absolutely true. Then, there's "stable", which means no loss (with
> > respect to acknowledged file ops, ala NFS stability rules). You've
> > rather forcefully made my point. Not really worth quibiling about
> > which kind of loss is acceptable or unacceptable.
> If you consider it quibbling, _you_ have missed _my_ point, which is
> that yes, some people consider some kinds of loss risk acceptable while
> considering other kinds of loss risk unacceptable. For example, I
> consider my first example ("if you lose power the file being written
> may have garbage contents") acceptable; I do not consider my third
> example ("if you lose power you mgiht as well re-newfs") acceptable.
> > The goal, really, is commercially accepted standard of stability.
> "The" goal?
> *Your* goal, perhaps. Others' goals may well be - and in many cases
> doubtless are - different. (I also disagree with your implicit
> assertion that the level of stability specified for an NFS server is
> the "commercially accepted standard of stability", but that's really
> incidental to my principal point here.)
Fair enough. I do have an NFS-centric take on things.
Does FFS+softdep have acceptable loss characteristics
while ext3 has unacceptable loss characteristics? "Some people"
may find the ext3 characteristics "acceptable". And, really, that's
the quible I wanted to provoke before somebody argued that ext3 only
surpassed FFS on the postmark test by risking unacceptable loss.
That -- as you point out -- is in the eye of the beholder.
I'm not suggesting ext3 has greater loss risk than FFS,
or vice versa. Rather, that is moot.
The mount options for the original post are relavent to
loss risk and to performance.
> Of course, if you believe you have a filesystem that provides all these
> things (speed, stability, light weight, simple code) at once, please,
> make that clear, and contribute the code!
Ya, I'm afraid I can't do that.
However, my posting about a method for hashed directories
will make a huge improvement on postmark results.
And that's what the initiator of this thread is after.
> /~\ The ASCII der Mouse
> \ / Ribbon Campaign
> X Against HTML firstname.lastname@example.org
> / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B