Sounds to me then that reading volume local on server would be an
improvement, since RX would be out of the picture.
-- Nathan
> -----Original Message-----
> From: Nickolai Zeldovich [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, March 14, 2001 8:33 AM
> To: [EMAIL PROTECTED]
> Cc: [EMAIL PROTECTED]
> Subject: Re: [OpenAFS] OpenAFS Project List
>
>
> > I seem to recall that results of some performance tests
> were posted here
> > or on openafs-devel in the not-too-distant past. The
> performance problems
> > with vos dump appear to be primarily in the dump process
> itself, _not_ in
> > the network transfer.
>
> FWIW, I was also playing around with vos dump a while back and it
> seemed to me rx performance was the bottleneck. Trying to read the
> vnodes ahead of time in a separate thread, using aio to fetch vnodes
> in parallel, etc, didn't seem to help much at all, but increasing
> some rx parameters (eg number of packets between acks) seemed to
> get the "vos dump" performance up a bit.
>
> -- kolya
> _______________________________________________
> OpenAFS-info mailing list
> [EMAIL PROTECTED]
> https://lists.openafs.org/mailman/listinfo.cgi/openafs-info
>
_______________________________________________
OpenAFS-info mailing list
[EMAIL PROTECTED]
https://lists.openafs.org/mailman/listinfo.cgi/openafs-info
