monotone-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Monotone-devel] slow update/status on NFS


From: Richard Levitte
Subject: Re: [Monotone-devel] slow update/status on NFS
Date: Wed, 23 May 2007 11:51:30 +0200 (CEST)

In message <address@hidden> on Wed, 23 May 2007 10:26:53 +0200, Markus 
Schiltknecht <address@hidden> said:

markus> Hi,
markus> 
markus> Richard Levitte wrote:
markus> > I just looked at the patch, and although I can see that it
markus> > fixes things for status, I think it could be taken one step
markus> > further (and thereby generalising the solution) by modifying
markus> > the file_path class to have a status cache.
markus> 
markus> Shouldn't the OS do that sort of caching? That's probably why
markus> there's no difference with local filesystems. But why does the
markus> kernel not cache NFS stat() calls? Can we somehow teach the
markus> kernel to do that caching, so we don't have to write our own?

Yes, one could argue that the kernel "should".  As for NFS, there are
a lot of crap implementations (arguably, NFS itself is crap, so...),
but if we get back to the kernel, which kernel are you talking about?
I'm sure some kernels have calls with which you can tune this kind of
thing, but does it look the same everywhere?  How about in Winblows?

So, this is a tradeoff decision.  Trust all kernels or none, or figure
out what kernels need som extra help?

Cheers,
Richard

-- 
Richard Levitte                         address@hidden
                                        http://richard.levitte.org/

"When I became a man I put away childish things, including
 the fear of childishness and the desire to be very grown up."
                                                -- C.S. Lewis




reply via email to

[Prev in Thread] Current Thread [Next in Thread]