 |
 |
Hi, I see you have a 3 tiers hierarchy: redirector, supervisors and workernodes. What is a supervisor? Other configurations don't seem to have this layer. Can supervisors be specialised worker nodes? What is olbd? and finally is the software for redirector supervisors and data servers the same (i.e. I install the same rpm) but configured differently? Why are you using 2 xrootd instances per WN? Have you bound each with a different partition? Or do they both serve all the available partitions. Apologies for all these questions you'll have to bear with me until I get up o speed. thanks cheers alessandra Pavel Jakl wrote: > Hello, > > Yes, it is. But the plot is showing sum of usage at production and > development cluster. I have to split the statistics into 2 plots, sorry > about not giving you a proper plot. > I can say, that I couldn't see what Fabrizio is talking about because we > still don't have so many users using the xrootd. We wanted to move from > our old solution very slowly and see the behavior of xrootd with > increasing number of users as time goes. Users are switching from rootd > to xrootd by their choice and I can say that this process in completely > unclear to me :-) But the number of users is higher and higher as > xrootd gives more advantages. > > Cheers > Pavel > > Alessandra Forti wrote: >> Hi Pavel, >> >> thanks this is very useful. Is this the STAR production cluster? I >> read also that talk given at chep. >> >> cheers >> alessandra >> >> >> Pavel Jakl wrote: >>> Hello Alessandra, >>> >>> Andy talked mostly about scalability and performance tests of xrootd >>> and trying to show the behavior of xrootd at upper limits I believe. >>> I don't know nothing about the fact that xrootd uses a lots of memory. >>> We are using xrootd on WNs and we have deployed it on 320 nodes with >>> about 130TB. These nodes are heavily loaded by user's jobs, mostly >>> ROOT jobs. >>> For your imagination, I am sending you plots being produced by >>> really home-made monitoring with Ganglia toolkit. >>> I also need to mention, that you have to divide these numbers by the >>> factor of two, because we are running two instances of xrootd per >>> node (we have a development and production cluster). You can see that >>> memory usage is really, really small. >>> >>> Cheers >>> Pavel >>> >>> Alessandra Forti wrote: >>>> Apologies that worries me in using xrootd on the WNs.... >>>> I'm a bit confused right now :( >>>> >>>> Alessandra Forti wrote: >>>>> Hi Peter, >>>>> >>>>> after yesterday I also remember that the second thing that worried >>>>> me in using dcache on the WN is a comment Andy made about the use >>>>> of memory in his talk at CHEP I believe. He said it xrootd uses a >>>>> lot of memory but there weren't number specified. >>>>> >>>>> My system is double cpu with 2GB of memory per cpu. Considering >>>>> that an atlas job can use more than 1 GB we are now at 1.1 I thik. >>>>> Will it be enough? I think so but I just wanted to check. >>>>> >>>>> cheers >>>>> alessandra >>>>> >>>>> >>>>> Peter Elmer wrote: >>>>>> Hi All, >>>>>> >>>>>> There is now a new xrootd development version: xrootd >>>>>> 20060523-1741, please see: >>>>>> http://xrootd.slac.stanford.edu/download/20060523-1741/ >>>>>> >>>>>> for downloads. >>>>>> >>>>>> Relative to the last development build (20060418-0404) this >>>>>> includes a variety of small bug fixes, plus one important one for >>>>>> the redirector. See >>>>>> the xrootd.History file for more details. I've included a link to the >>>>>> SL3 debuginfo rpm on the rpm page. (Although I've not tried it >>>>>> myself, >>>>>> so I have no idea if it works! Feedback is welcome.) >>>>>> >>>>>> Gerri, if it is still possible, you could add this to the next ROOT >>>>>> build, too. >>>>>> >>>>>> For the full set of changes and links to rpms/tarballs to >>>>>> download see the >>>>>> the xrootd web page and/or version history: >>>>>> >>>>>> http://xrootd.slac.stanford.edu >>>>>> http://xrootd.slac.stanford.edu/xrootd.History >>>>>> >>>>>> Let us know if there are problems. >>>>>> >>>>>> Pete >>>>>> >>>>>> ------------------------------------------------------------------------- >>>>>> >>>>>> Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) >>>>>> 767-4644 >>>>>> Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, >>>>>> Switzerland >>>>>> ------------------------------------------------------------------------- >>>>>> >>>>> >>>> >>> >>> >>> ------------------------------------------------------------------------ >>> >>> >>> ------------------------------------------------------------------------ >>> >>> >>> ------------------------------------------------------------------------ >>> >>> >>> ------------------------------------------------------------------------ >>> >> > -- ******************************************* * Dr Alessandra Forti * * Technical Coordinator - NorthGrid Tier2 * * http://www.hep.man.ac.uk/u/aforti * *******************************************