Messages in this thread | | | Date | Mon, 23 Mar 1998 21:24:25 -0500 (EST) | From | "Steven N. Hirsch" <> | Subject | Re: Why is NFS in 2.0.3? slow |
| |
On Mon, 23 Mar 1998, David S. Miller wrote:
> Date: Mon, 23 Mar 1998 16:32:29 -0800 > From: Jason Venner <jason@idiom.com> > > I get 500k/sec for NFS traffic. This really sucks. > Doesn't seem to matter Read or Write. > I don't do TCP/IP. I don't set any block sizes, just use defaults on > the mounts. > > That is pretty horrific, I do agree. I remember that when Olaf first > coded up the new NFS code he was able to get ~4MB/s NFS in the lab of > Linux boxes last year at Linux Expo. Whether he tuned parameters > etc. to get that, I don't know. If he did, we need to investigate our > default parameters a bit to see if we can find more suitable settings.
I can get throughput in the neighborhood of 4MB/sec. between my Alpha UDB and a P150+ over 100-Base ethernet. Both are running 2.1.90 kernels and fitted with Netgear FA310TX "tulip" network adapters.
However, there is room for improvement. At work, I administer a cluster of AlphaStation 600/333 boxes linked with 100Base, and have measured almost 10MB/sec. copying large files between them! Granted, we're talking expensive machines, Digital Unix 4.0 and NFS-v3 but, hey, it's something to shoot for <g>.
Steve
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |