Personally, I'm a big fan of packet traces. They can tell you alot. In this case, you have one good performing client, and one bad one. Take a packet trace of each and compare them. What sizes are the clients requesting? What calls are they using? Is the client just not generating requests very quickly?

Just some thoughts..

-- Adam Fox NetApp Professional Services, NC adamfox@netapp.com

-----Original Message----- From: Link, Bill (OTS-EDH) [mailto:Bill_Link@billing.com] Sent: Thursday, January 04, 2001 4:57 PM To: 'toasters@mathworks.com' Subject: Slooow SGI Challenge

Has anyone had any experience with netapp on an SGI Challenge system?

I have an F760 attached through a cisco 4000 switch to an sgi challenge and an IBM S7A. I have this arrangement on a dedicated subnet with all interfaces set manually to 100tx. (The vme controller in the SGI will not abide full duplex). The netapp is running 5.3.6R1, the SGI is running IRIX 6.5.5, and the RS6K is running AIX4.3.3.

The IBM will create a 128M file in under 30 sec. If I mount the filer on the SGI box and specify tcp it takes about 2 min 45 sec to write the same file. If I let the mount on the SGI default to udp it won't write more that 15 or 20 mb before it starts getting nfs timeouts. The filer seems oblivious to any trouble that the sgi is having. The sgi will work (slowly) at 10mb half duplex with udp.

Any suggestions would be much appreciated.

Bill Link SysAdmin Output Technology Solutions bill_link@billing.com