Since the switch is not a managed switch, set your
server nic to auto. If that does not work drop the nic to 100/full and see what
you get. If the throughput is higher at the 100/full it will be either a driver
on the server side or the switch. I recommend getting a managed switch. when
your running gig you want to be able to see both sides.
Ed
Actually I
know the cables are OK because I have used the intel drivers to validate the
cables. All indications (lights on cards, and switch), along with diagnostics in
the drivers, all indicate gigabit full duplex
I am not using a managed
switch, so I can't get any info from it.
I have also used iPerf which
shows me similar results.
-Scott
----- Original Message
-----
From: "Marc Luethi" <netztier@xxxxxxxxxx>
To: "Community
support list for Wireshark" <wireshark-users@xxxxxxxxxxxxx>
Cc: "Scott
Chapman" <WireShark@xxxxxxxxxxxxxxxxx>
Sent: Saturday, March 7, 2009
4:14:13 AM GMT -05:00 US/Canada Eastern
Subject: Re: [Wireshark-users] Slow
gigabit network
On Fri, 2009-03-06 at 21:29 -0500, Scott Chapman
wrote:
> So, the problem I have is that I get about 12-14MB/sec when I
copy
> files.
Might be protocol overhead, as (almost) usual. Yet,
assuming that the
cabling is okay from the fact alone that both cards connect
at 1Gbps is
wrong. There might still be CRC errors (and hence packet loss and
TCP
retransmissions), or flow control might be stepping in:
http://www.smallnetbuilder.com/content/view/30212/54/
If the
switch is a managed one, have a good look at each port counter,
or see if you
can undig some of those counters from the depths of your
NIC's driver.
Wireshark might even be able to show flow control pause
frames being sent or
received by the system it runs on.
Get a Windows or Java version of iPerf
or jPerf, and use it to shove
data from box to box. Here's how - same
procedure for Linux and
Windows.
http://ubuntuforums.org/showpost.php?p=6758444&postcount=7
http://ubuntuforums.org/showpost.php?p=6522634&postcount=4
If
you get reasonable throughput numbers here, you can safely assume
that the IP
stacks and storage I/O systems are in good shape.
Then you can start to
deep into the analysis of the involved protocols
like CIFS, NetBIOS-over-IP
or NFS.
regards
Marc
