On Mon, Aug 25, 2008 at 12:40 PM, John Bresnahan <[EMAIL PROTECTED]> wrote:
>> Those options did the trick.  Thanks!
>
> was it -con 1 that made the difference or tcp-bs?

The "tcp-bs" option set to the 2MB you suggested seems to be the magic
number.  I toggled it up and down (trying various contortions of BDP)
along with the number of parallels and, for the most part, performance
would start to tank.

>
>
>>  Another quick question: Is
>> there an easy way to get performance stats from RFT?
>
> there is not but perhaps that is a feature which could be added.  can you
> tell me what specifically you are looking for?

I was mainly hoping to determine the final throughput rate so if there
was some way to get some brief/simple statistics much like the verbose
mode for globus-url-copy output that would be great.  Does the RFT
database keep that info?  I mean, can users make simple queries on it?

Reply via email to