perfsonar livecd offers npad service that remote hosts can connect and see the performance and results. http://www.internet2.edu/performance/toolkit/index.html TcpOptimizer helps tunning the tcp/ip for windows systems. http://www.speedguide.net/downloads.php nuttcp is good to generate packets/sec. -Azher Nathan Ward wrote:

On 3/11/2009, at 10:56 AM, Mark Urbach wrote:

...

Anyone have a good solution to get "accurate" speed results when testing at 10/100/1000 Ethernet speeds?

If you want accuracy, you want to buy a packet generator/router tester unit.

I just built a tool for a customer (a last-mile network provider) that runs a series of iperf tests over several days, and generates a report. iperf works well enough, but it seems to be much better when driven by humans, vs. driven by scripts.

I'm not aware of any free tools that do just ethernet frames.

...

Do you have a server/software that customer can test too?

Not sure what you're after here - do you want to host your own speedtest.net-like service so your customers can self-test their access links? Does this mean much, or should they be testing against a server outside your network?Also, if you host your own service and you're talking about 10/100/1000mbit connections, you might want to put something in place that prevents several people testing at once.

-- Nathan Ward

-Azher

On Mon, Nov 02, 2009 at 01:30:18PM -1000, Michael Painter wrote:

Nathan Ward wrote:

...

On 3/11/2009, at 10:56 AM, Mark Urbach wrote:

...

Anyone have a good solution to get "accurate" speed results when testing at 10/100/1000 Ethernet speeds?

An NDT server?... such as: http://ndt.anl.gov:7123/

I just tested that server, and couldn't get any results which were even vaguely close to accurate. Of course it probably didn't help that the only routes I could find to the test server were either Chicago - Palo Alto - Chicago or Chicago - Ashburn - Chicago, but this doesn't seem like it would ever be useful for testing gigabit anything. For end user testing, I've actually seen reasonable results from speedtest.net. http://www.speedtest.net/result/610596179.png for example, better than ndt.anl.gov at any rate. :P For quick and dirty high speed Internet testing up to a gigabit, this is my favorite standby (it often helps to eliminate your local disk from the equation by writing the downloaded file to /dev/null too):

fetch -o /dev/null http://cachefly.cachefly.net/100mb.test /dev/null 100% of 100 MB 102 MBps

But the best (and conveniently enough the most commonly used) tool for in-depth high speed testing was already mentioned, iperf. Another useful tool if you're trying to troubleshoot tcp issues is http://www.tcptrace.org/. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)

Date: Mon, 2 Nov 2009 18:22:19 -0500 From: Jon Meek <meekjt@gmail.com>

I use iperf with packet capture on both sides, then analyze the packet capture for per-second throughput and re-transmits. I usually do 10 TCP streams for 30 seconds.

Note that on GigE with significant RTTs (5-15 ms) some TCP tuning is needed to deal with the bandwidth delay product. It is also possible that Ethernet drivers will have an effect. Local testing of the pair of test machines should be done if you can't get to about 980 Mbps on a Gig link (keeping in mind the comment about TCP tuning as latency increases).

Jon

On Mon, Nov 2, 2009 at 4:56 PM, Mark Urbach <mark.urbach@pnpt.com> wrote:

...

Anyone have a good solution to get "accurate" speed results when testing at 10/100/1000 Ethernet speeds?

Do you have a server/software that customer can test too?

I'll also suggest http://fasterdata.es.net as a resource for network tuning. Tuning TCP is hard. UDP is simple, but some things can even impact UDP. Many less than obvious things can have a huge impact on high-speed data transfer. The choice of congestion algorithms can be very significant. As anyone who has used bittorrent should have noticed, having multiple TCP streams works better than a single stream. An oddity we have noted is that some routers will process switch layer 2 traffic if a layer 3 interface exists on the port even if it is unconfigured and unused. Man, that kills performance, even in low latency situations! FWIW, we use mostly iperf, but may be biased as the iperf maintainer works here. We did start using iperf before we hired him, though. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634 Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751

Please take note with using iperf that you'll want to make sure the appropriate TCP Window Size has been negotiated. We recently did some testing with systems that had decided to pick less than optimal window sizes and in turn had to manually set the size within iperf options. Jason On Tue, Nov 3, 2009 at 4:01 AM, Benoit VANNIER <benoit.vannier@apog.net>wrote:

Hello,

Iperf is pretty good at this ... It s free

Ben

-----Message d'origine----- De : Mark Urbach [mailto:mark.urbach@pnpt.com] Envoyé : lundi 2 novembre 2009 22:57 À : nanog@nanog.org Objet : Speed Testing and Throughput testing

Anyone have a good solution to get "accurate" speed results when testing at 10/100/1000 Ethernet speeds?

Do you have a server/software that customer can test too?

Thanks, Mark Urbach PinPoint Communications, Inc. 100 N. 12th St Suite 500 Lincoln, NE 68508 402-438-6211 ext 1923 Office 402-660-7982 Cell mark.urbach@pnpt.com [cid:image003.jpg@01CA5BD5.1A5CEE20]

-- Jason Biel

Linux always worked best for us as well, was easy running a livecd with laptops. We found that two windows XP machines, same identical hardware and OS load yielded different registry settings (or lack thereof) for TCP Window setting. Jason On Tue, Nov 3, 2009 at 6:18 AM, Bret Clark <bclark@spectraaccess.com> wrote:

True, we usually find Linux based machines work better running IPerf then Windows (at least out of the box) because of the TCP window size....well Windows XP at least, don't know about Vista or 7. Jason Biel wrote:

Please take note with using iperf that you'll want to make sure the appropriate TCP Window Size has been negotiated. We recently did some testing with systems that had decided to pick less than optimal window sizes and in turn had to manually set the size within iperf options.

Jason

On Tue, Nov 3, 2009 at 4:01 AM, Benoit VANNIER [1]<benoit.vannier@apog.net

...

wrote :

Hello,

Iperf is pretty good at this ... It s free

Ben

-----Message d'origine----- De : Mark Urbach [[2]mailto:mark.urbach@pnpt.com] Envoyé : lundi 2 novembre 2009 22:57 À : [3]nanog@nanog.org Objet : Speed Testing and Throughput testing

Anyone have a good solution to get "accurate" speed results when testing at 10/100/1000 Ethernet speeds?

Do you have a server/software that customer can test too?

Thanks, Mark Urbach PinPoint Communications, Inc. 100 N. 12th St Suite 500 Lincoln, NE 68508 402-438-6211 ext 1923 Office 402-660-7982 Cell [4]mark.urbach@pnpt.com [[5]cid:image003.jpg@01CA5BD5.1A5CEE20]

References

1. mailto:benoit.vannier@apog.net 2. mailto:mark.urbach@pnpt.com 3. mailto:nanog@nanog.org 4. mailto:mark.urbach@pnpt.com 5. cid:image003.jpg@01CA5BD5.1A5CEE20

-- Jason Biel

We actually used both tests to finally narrow down the TCP window size issue. We first used the UDP test to make sure the link was good. Jason On Wed, Nov 4, 2009 at 10:54 PM, Simon Horman <horms@verge.net.au> wrote:

On Tue, Nov 03, 2009 at 06:49:05AM -0600, Jason Biel wrote:

...

Linux always worked best for us as well, was easy running a livecd with laptops. We found that two windows XP machines, same identical hardware and OS load yielded different registry settings (or lack thereof) for TCP Window setting.

I would have thought that (netperf's) UDP tests were more appropriate than TCP for testing link speed. Am I missing the point?

-- Jason Biel

ml@axmo12.de wrote:

Hi Michael,

Zitat von Michael Helmeste <mhelmest@uvic.ca>:

...

[...] You can give it a try at: http://caranthir.dao.nrc.ca/netperf-www/ (login nanog/nanog). User guide here: http://www.cadc-ccda.hia-iha.nrc-cnrc.gc.ca/netperf/testdetail.shtml

I might end up packaging and releasing the code if there is interest.

If you have time it would be great if you could do that. - I'm very interested in it! +1