Messages in this thread | | | Date | Mon, 12 Apr 1999 10:56:52 -0400 (EDT) | From | Andrew Schretter <> | Subject | Re: NFSv3 client for Linux-2.2.5 ready for alpha testing... |
| |
Eric, Sorry I was so vague. I understand what you are saying, but I doubt the disk speed of the server is the major bottleneck (for one thing I used an Ultra 60 that was not serving anything except the private export used for testing.) The problems I have been encountering (for almost 2 years now) can be summarized as follows (reproducible results).
These tests were done on a 100Mb switched network (Cisco 5000 switch) with minimal network/processor load on the systems.
S1 = Sun Ultra 60 Solaris 2.6 100Mb Ethernet S2 = Sun Ultra 60 Solaris 2.6 100Mb Ethernet S3 = Sun Ultra 1 140Mhz Solaris 2.6 10Mb Ethernet L1 = Linux 450 Pentium II 100Mb Ethernet NFSV2 L2 = Linux 450 Pentium II 100Mb Ethernet NFSV3 ...Speeds are reported Read/Write in megabytes per second...
CLIENTS S1 S2 S3 L1 L2 S E S1 ** 8.0/7.5 0.8/0.7 7.0/0.5 0.5/4.5 R V L1 8.0/7.5 8.0/7.5 0.8/0.8 ** 8.0/7.0 E R
As you can see, performance is quite nice except for a few isolated cases.
You'll notice that a linux server is the way to go (except that I need locking so I need knfsd and I'm still experimenting with it for stability).
I have a much more detailed summary I can send you but this is the basics of it.
In short, my network can read/write very fast Solaris -> Solaris and Linux -> Linux, but NFSV2 Linux writes slow to Solaris and NFSV3 Linux reads slow from Solaris.
Andrew
On Mon, 12 Apr 1999, Eric Werme USG wrote: > > Andrew Schretter wrote: > > > > FINALLY, I can write from a linux machine via nfs to a Solaris server > > at right around 2 megabytes per second (100MB Lan). > > Huh? > > I don't get this.. Everyone but me appears to have problems writing to > Solaris NFS server! > > Also, the fs that it is writing to is a raid0 array spread across 18 > drives.. > > The synchronous writes on NFS V2 "require" that the data be safely on disk > before a reply is made. Even the V3 clients with unstable writes should? > must? had better do a COMMIT at close time to catch any errors, including > out-of-space errors the biods haven't reported yet. > > Therefore, NFS writes are limited to disk speed, and V2 writes suffer from > several seeks per write unless assisted by write gathering, NVRAM, and > related whatnot. > > I would not be surprised if most people on this list don't have an 18 > disk array to help spred the load. > > BTW, given your hardware, 1963 KB/sec is not terribly exciting. V3 should > easily see that with a single disk. What do you get from some other V3 > client? Expect that Trond will be pestering you for data ASAP. > > -Ric Werme > > - > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.rutgers.edu > Please read the FAQ at http://www.tux.org/lkml/ >
Andrew Schretter Systems Programmer, Duke University Dept. of Mathematics (919) 660-2866
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |