Skip to Content.
Sympa Menu

ndt-users - RE: Adapting NDT for parallel TCP flows

Subject: ndt-users list created

List archive

RE: Adapting NDT for parallel TCP flows


Chronological Thread 
  • From: "Webb, Ken" <>
  • To: <>
  • Subject: RE: Adapting NDT for parallel TCP flows
  • Date: Thu, 17 Apr 2008 19:01:13 +0200

Thanks Richard,

Because I'm in a hurry I probably will go for something based on BWCTL
for now: thanks for the pointer. But it seems like an interesting
project so I hope to return to this in future. I will of course send you
any changes I make.

Regards,

Ken

-----Original Message-----
From: Richard Carlson
[mailto:]

Sent: Thursday, April 17, 2008 3:11 PM
To: Webb, Ken;

Subject: Re: Adapting NDT for parallel TCP flows

Hi Ken;

Thanks for the enquiry. At the present time I don't know of anyone
working on a parallel flow module.

The NDT system was designed to look for a very specific set of known
problems that impacted performance. The idea is that the campus NOC
runs a local server and when the phone call comes in saying "what's
wrong with the network?" the NOC staff gives the caller the NDT URL
and says, "please run this test and we'll look at the results". This
lets the community get real data on possible problems and we can get
past the finger pointing stage we usually get stuck in.

Thus the NDT was not designed as a generic test infrastructure or as
a replacement for something like Iperf. However, we have had some
interest in expanding the NDT system to support more functions. For
example the folks at NCMC extended the system to allow the client to
automatically repeat tests so they could monitor multiple links
without having to install HW/SW at remote sites. Once this code was
working they shared the mods with us and we incorporated that
functionally into the current distribution. It's a sysadmin option
that can be enabled by adding some PARAM options into the NDT web page.

Right now I have no plans to implement parallel TCP streams, and
would point someone to BWCTL for this function. However, I would
have no objection to your checking into how to modify the NDT source
code if you want/need this functionality. All I ask is that you send
me the patches that I can put into the NDT contrib directory or
incorporate into the main NDT distribution.

Regards;
Rich
NDT developer

At 05:44 AM 4/17/2008, Webb, Ken wrote:
>Hi,
>
>I was wondering whether anyone has tried to adapt the NDT client to run
>parallel TCP sessions. The server I believe should already handle this,
>although I would still like to be able to limit the server to one
client
>host at a time.
>
>In our test environment I'm seeing the sort of 'random' packet loss
>which seems to favour multiple parallel flows (with small window size)
>over single flows. As I'm interested in periodically measuring the
>effective size of the pipe available I'd like to adapt the client to
run
>multiple sessions. Has anyone tried or does it run contrary to the
>philosophy and design of NDT?
>
>Ken Webb,
>UPC Broadband,
>Boeing Avenue 53,
>1119PE Schiphol Rijk,
>The Netherlands.

------------------------------------



Richard A. Carlson e-mail:

Network Engineer phone: (734) 352-7043
Internet2 fax: (734) 913-4255
1000 Oakbrook Dr; Suite 300
Ann Arbor, MI 48104




Archive powered by MHonArc 2.6.16.

Top of Page