 | | | nfs/network latency | nfs/network latency 2004-10-14 - By erik.welch@(protected)
Back
I have 2 Cisco switches A & B.
A is connected to a NetApp (gigabit).
B & A are connected via fibre to a router (in another bldg).
If any of our opterons (Sun, HP, or Dell) are hooked up to switch A we
have horrible network throughput (1MB/sec) and piles of nfs timeouts. If
they are hooked up to switch B the throughput is good(10MB/sec), even
though it has to go through a convoluted path. I don 't see errors
anywhere: switches, linux boxes, NetApps, or on sniffer.
BTW, I am defining network throughput as speed at which I can copy data
from NetApp to system.
My guess is that switch A is breaking the packets into more manageable
chunks.
One of our network team suspects it has to do with the IGP (Inter-Packet
Gap) on the RedHat boxes. But that wouldn 't affect xmit from NetApp to
Linux, would it? Also, no other system types (HPUX, Solaris, W2K, etc)
have this problem when hooked to Switch A.
Anyone know how to set the igp for a tg3 (Tigon) module?
Any other ideas as to the cause? Solutions?
erik
----
Erik Welch Ph: 408-474-5362 Cell:
408-690-2698
Unix & Backups System Administrator Philips Semiconductor
erik.welch@(protected)
"Axiom: Computers are cheap, fast, and stable. Pick two "
--
Taroon-list mailing list
Taroon-list@(protected)
http://www.redhat.com/mailman/listinfo/taroon-list
|
|
|