 |
 |
Hi, > You are right I am involved in Atlas production and using dcache storage > element. > > I thought dcache with TSM (or HSM) will be useful for babar as well. > > I never realised that xrootd also interfaced with mass storage system. > We can try this as well. Can you point me link which gives details about it ? > > Do you know any group using dcache with TSM in our babar group ? I did such tests with TSM in the past and I'm trying to improve those now that I swithed to CMS. I sent Asok the material/info I already have. Regards, -- Gregory >> hello Ashok, >> >> what is the reason for choosing dcache instead of >> xrootd ? (are you involved also in LCG ?) >> the latter can also be interfaced with a mass storage system. >> cheers, >> JY >> >> [log in to unmask] wrote: >>> Hi Wilko, >>> >>> Are you using dcache xrdxp command ? Do you use dcache for storing >>> the >>> data ? >>> >>> At Uvic, we are installing dcache with tape (TSM) to store BaBaR data. >>> We can copy the data to tape and access it through xrootd door. >>> I am wondering how to keep bookkepping about it ? >>> >>> Any advise ? Do you know anybody storing BaBar data on dcache and >>> accessing throuch xroot door and keeping bookkeeping of it ? >>> >>> Thanks, >>> >>> Ashok >>> >>> >>> >>>> Hello Fabrizio >>>> >>>> I will start testing xrootd and xrdcp again (I was gone for a couple of >>>> weeks). >>>> I will do some larger test with xrdcp as I would like to try using >>>> it for the Babar data distribution. >>>> >>>> Cheers, >>>> Wilko >>>> >>>> >>>> >>>> >>>> >>>> On Mon, 18 Jun 2007, Fabrizio Furano wrote: >>>> >>>> >>>>> Hi all, >>>>> >>>>> it looks like I finished this small dev/bugfix round on XrdClient. >>>>> Very >>>>> few >>>>> added features: >>>>> - function to clean the cache forcefully >>>>> - windowsize agreement between client and server in multistream mode >>>>> (i.e. >>>>> you only need to set your preferred WAN tcp windowsize on the server >>>>> side) >>>>> - TestXrdClient_read now has the --check option which does a basic >>>>> correctness check on the data it receives. >>>>> >>>>> For the bugfix part, the most important ones are: >>>>> - correct interpretation of error messages coming out as unsolicited >>>>> responses to staging requests (needs validation from Andreas) >>>>> - correct behavior of the cache with 'nasty' chunk requests >>>>> - correct optimization of concurrent connection creations >>>>> - mem corruption in creating connections in the connection manager >>>>> - mem corruption in purging unused connections in the error recovery >>>>> mechanism in the multistream case >>>>> >>>>> Right now I am finishing the tests, but everything looks fine up to >>>>> now. >>>>> I'd >>>>> like to hear from you if you have problems with the latest version. >>>>> >>>>> BTW xrdcp with the latest code is scoring 6.1MB/s between PD and SLAC, >>>>> against the previous 4.5-5MB/s. It would be nice to hear from you if >>>>> you >>>>> confirm this improvement or if it's just in my setup for some strange >>>>> reason. >>>>> >>>>> Fabrizio >>>>> >>>>> >>> >>> >>> >> >> > > >