ndt-users list created

["Ken Bellars" <>]

Rich,

Thanks so much for walking me through building my own NDT. The first suggestion actually fixed the problem. 

Now its running cool but I think there is still a problem (most likely the last.).

My 
C2S speed is far greater than the S2C. See below.

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

click START to re-test
Connected to: 196.207.6.22  --  Using IPv4 address
Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Done
checking for firewalls . . . . . . . . . . . . . . . . . . .  Done

running 10s outbound test ( Client -to- Server [C2S] ) . . . . . 44.25Mb/s
running 10s inbound test ( Server -to- Client [S2C] ) . . . . . . 1.48Mb/s

The slowest link in the end-to-end path is a 10 Gbps 10 Gigabit Ethernet/OC-192 subnet
Information: Other network traffic is congesting the link
[ S2C ]: Excessive packet queuing detected

click START to re-test

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

Now I know it has to do with the 'etc/sysctl.conf' but i can't think of the solution because I was using the sample you sent to me.

I will appreciate your input in this regard.

Regards,
Kenny

On 10/17/06, Richard Carlson <> wrote:

Hi Kenny;

There is a bug in the Web100 user library (libweb100) that the NDT server
process hits.  To see if this is the case, add the "-4" option to the
web100srv process.  You can modify the /etc/init.d/ndt script to change the
WEB100SRV_OPTIONS line or manually kill/restart the web100srv process.

The other reason could be a hung web100srv process.  I introduced a bug in
the client code that MAC and BSD hosts using the command line client can
trigger.  I've fixed the client code and am looking for a server side fix
now.  Send me the output from the "netstat -nat" command if the above
action doesn't fix this problem.

Rich

At 03:04 AM 10/17/2006, Ken Bellars wrote:
>Richard,
>
>I have done the above and now what Ii keep getting is
>
>-------------------------------------------------------------
>Checking for Middleboxes . . . . . . . . . . . . . . . . . .  Protocol error!
>-------------------------------------------------------------
>
>Is any wrong on my side?
>
>Thanks
>Kenny
>
>
>On 10/13/06, Richard Carlson <
><mailto:> > wrote:
>Hi Kenny;
>
>Here's a copy of the stuff I add to my /etc/sysctl.conf file.
>
># NEW parameter for 2.6.13 kernels.  Set to Reno style CC
>net.ipv4.tcp_congestion_control = reno
>
># TCP Tuning data added 7/28/05 RAC
># increase Linux TCP buffer limits
>net.core.rmem_max = 33554432
>net.core.wmem_max = 33554432
>net.core.rmem_default = 65535
>net.core.wmem_default = 65535
>
># increase Linux autotuning TCP buffer limits
>net.ipv4.tcp_rmem = 4096 87380 16777216
>net.ipv4.tcp_wmem = 4096 65535 16777216
># net.ipv4.tcp_mem = 1024 65535 524288
>
># Turn on Explicit Congestion Notification (ECN) processing
>net.ipv4.tcp_ecn = 1
>
>A few things to note.
>
>The INSTALL file that comes with the NDT package is suppose to have this
>information.  Unfortunately the current tarball contains a generic GNU
>based INSTALL file.  The NDT specific version has been returned and future
>tarballs will contain this tuning info.
>
>Also note that the net.ipv4.tcp_mem settings are commented out.  Looking at
>the Linux documentation it appears that the kernel does a better job of
>setting the right values than manual intervention.  I recommend letting the
>kernel handle this.
>
>One last thing, the latest Linux 2.6.x kernels do autotuning.  Thus these
>tuning values try and set both reasonable default (64KB) and max limits (16
>- 32 MB).  This autotuning can be turned off by using the NDT "-b value"
>option.  I don't recommend using this option, but it is there in case you
>need it.
>
>To learn more about this follow the links on the main NDT web page to the
>PSC and LBNL tuning pages.  Or google "tcp tuning" to find these pages.
>
>Rich
>
>At 06:32 AM 10/13/2006, Ken Bellars wrote:
>
>
> >I have my NDT running perfectly now.
> >
> >I however noticed that editing the config file sysctl.conf   affects the
> >report.
> >
> >I need to know the best config for the more close to accrate report
> >
> >Thank you.
> >
> >Kenny
>
>------------------------------------
>
>
>
>Richard A. Carlson                              e-mail:
><mailto:>
>Network Engineer                                phone:  (734) 352-7043
>Internet2                                       fax:    (734) 913-4255
>1000 Oakbrook Dr; Suite 300
>Ann Arbor,  MI  48104
>

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



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