I also recorded the whole meeting for those who are keen to reply the experience (I will send that to jens to upload to the agenda page).

We spent the whole main meeting on item 1:
Atlas sonar test show poor data rates for UK sites with the possible exception of QMUL.
The conclusions were: 
- brians study make it seem unlikely that it is network capacity that is the issue. 
- qmul have 1 gridftp server, different max tcp settings, lustre , no firewall 
- manchester also have no firewall so thats unlikely to be the issue. 
- it could be something with DPM/ gridftp 
- Sam (and maybe other sites) will try different tcp settings for future tests (tests are run Mon-Thu). 
- We will also look into the history of test results (if poss)  and whether those performing badly are just doing more transfers at the time etc...

We will probably have to return to this next week (though maybe not for the whole meeting). 

After the main meeting we discussed briefly the workshop (specifically Dell participation): 
- we have an extra slot after the main meeting Wed ~3.30 - 5.30. 
- Dell are keen to come and happy to speak - they also want to come on Thu. 
- We decided to have a hw part 1 on wed which Dell will be encourage to come to and speak in; then general stuff on Thu morning which they can come to if they (really) want and then a hw pt2 in the afternoon in which we can discuss vendors and which we will say is closed. 

- We then discssued a bit Govinds db recover issues, Sam had a few tips which Govind will try and get back to us if he is still having problems.