DAST: Iperf - Huge difference between running iperf -c and iperf -s on Gigabit Ethernet - david --at-- lisas.de (fwd)
Contacting DAST re: Request for information about Iperf
From: David Vogler <david --at-- lisas.de>
Subject: Huge difference between running iperf -c and iperf -s on Gigabit
Ethernet
Question/Comment:
Hi iperf team,
me and my LUG don't have an answer for this strange behaviour. I recently
bought a Gigabit Ethernet PCI card for my desktop computer (cheap rtl8169
chipset). I have a file server with an onboard rtl8169 chipset and a
laptop with an Intel Pro 1000 chipset.
When measuring with iperf, I get approx 600 MBit/s between server and
laptop _both_ ways (running iperf -c server on laptop and iperf -s on
laptop). When running iperf -c server on the desktop to the server I get
only 250 MBit/s, running iperf -s on the desktop gets me 580 MBit/s
between server and desktop.
All three are running Fedora Core 6 i386, except the server which is a
FC6_AMD64 installation.
It's not a real problem, because my harddrives are slow enough ;), getting
50MB/s transfer from server to desktop via ftp is excactly the transfer
rate I get when copying from one server hardrive to another server
harddrive.
I just want to know, why iperf shows such a huge difference between
uploading from the desktop to the server, while downloading is fast.
I got the approx the same difference (20 MBit/s vs 8 MBit/s) when
connecting my laptop via WLAN (802.11g) to my network. Is this a known
problem with iperf? I didn't find it in the FAQ or via google and my
friends at my LUG experienced the same differences even with "better" PCI
cards, like the Intel Pro 1000.
With kind regards,
David
John S. Estabrook
jestabro --at-- ncsa.uiuc.edu