Re: [Nbd] Performance tuning of nbd !?
- To: nbd-general@lists.sourceforge.net,Paul Clements <paul.clements@...124...>
- Subject: Re: [Nbd] Performance tuning of nbd !?
- From: "Dirk von Suchodoletz" <dirk.von.suchodoletz@...65...>
- Date: Mon, 03 Mar 2008 23:40:53 +0100
- Message-id: <web-98279136@...164...>
- In-reply-to: <web-98178015@...164...>
- References: <Pine.LNX.4.64.0802282217000.5163@...163...> <47C83C11.8060401@...124...> <web-98178015@...164...>
Hi!
> > Yes. I think you should be able to use about 70% of the bandwidth (I can
> get
> > about 32Mb/s out of 45Mb/s (T3) with 100ms latency). You might need to
> > increase your TCP window size and/or change the I/O scheduler you're using
> > for NBD. Switch to "deadline" scheduler if you're not using that one
> already.
Some more background information: The used filesystem is squashfs with
blocksize of 64kByte. The read-ahead of NBD is set to 128kByte (seems to be the
optimal value there). I checked the thoughput of the rootfs (squashfs on NBD)
for a booting system. While I got up to 14000kbit/s for an ftp download (same
server), the nbd bandwidth usage never exceeded 4200kbit/s roughly one third of
the actual bandwidth available (independently of various variations of the tcp
buffer adjustments as suggested) :(
Any further ideas!? Dirk
Reply to: