Discussion list for thrulay, a network testing tool

["Cottrell, Les" <>]

Interesting. As I understand it from Connie, the client/sender (nettest2) is 
also being used to send to Caltech and CERN and there no such elongated RTTs 
are seen.  The nettest2 box is a Linux box, Intel Xeon with 2*3GHz cpus, 
windows ~ 20MB, 1GE NIC, Linux nettest2 2.4.21-27.8.ELsmp #1 SMP Fri Jan 14 
19:39:42 EST 2005 i686 i686 i386 GNU/Linux. 

-----Original Message-----
From: stanislav shalunov 
[mailto:]
 
Sent: Monday, March 28, 2005 5:41 PM
To: Cottrell, Les
Cc: Logg, Connie A.; 
;
 Dantong Yu; 

Subject: Re: 3 site comparison

"Cottrell, Les" 
<>
 writes:

> We ran pings while running the thrulay. The pings were pretty steady 
> at about 72ms, even when we ran the pings at 0.2 sec intervals.

Then this is internal queuing on the sender (=client).  (Is it perhaps a 
Linux box?)

> Dantong or Jason, are there any other machines we can test to at BNL 
> to see if we can isolate the problem?

That should help...

> Connie if you make measurements with thrulay from atlasprod3 then we 
> can see if all paths from atlasprod3 have similar RTTs.

Internal queuing on Linux happens.  Its significance is rather poorly 
understood and has little diagnostic value (at least by me and to me, at 
least so far).  Thrulay observes TCP's behavior like any bulk transfer would 
see it: multisecond delays.

-- 
Stanislav Shalunov              http://www.internet2.edu/~shalunov/

This message is designed to be viewed upside down.