Hi Flemming et al, On Sat, 13 Oct 2001 16:00:10 -0400 "Flemming Videbaek" <videbaek@sgs1.hirg.bnl.gov> wrote concerning "RCF Rcas nodes": > > The installation of this years increment to the RCF analysis farm > has dragged out., but has been ready use within the last week. The > addition is 12 new dual procesors each with ~60Gb of local disk- > this should expand the capabilities for post-reconstruction analysis > significantly. Cool! > In order to minimize the management and conflict of usage the new > rcf rcas machines as well as the older except for 0001-0004 will be > divided for use by specific Brahms collaborations institutions. The > current allocation is made with an eye for the present usage, as > well as active analysis, and has been circulated within members of > the the analysis group. Uh? How does allocating machines for specific BRAHMS institutes "minimize the management and conflict of usage"? In fact, I would argue that maintaince becomes a major hazzle when you have local dirs on various machines. What would really make it so much easier was if the ruptime application worked, so you could tell which machine had the least load and then lock onto that. Another thing would be to automatically put one onto the least busy machine, using some sort of software (I guess LSF is something like that). However, that would mean that the disks should not be mounted locally, but via NFS. Ok, so you may object to NFS, but if the machines are comparatively (physical) close - that is in the same building on the same switch, the overhead is minimal, especially with cheap 1GBit net avaliable. This may be a closed topic (though I've seen no discussion), in which case I'll crumple on privately. Another thing that could help, is if one runs rootd daemons on all the machines so that one can access files via TCP/IP from remote hosts (say the pii's). The man(1) rootd daemon page describes this. > The following > > rcas 0001-0004 > rcas 0005 Database server (may well move to anotehr one in future) Am I right in believing that you can no longer SSH to this machine? I think that's a good idea. > In addition one can utilize the lsf queues for distribution 'batch' > like jobs that do not use local disc resources to find machines with > the least load and utilization. Could you give a short summary of how to use LSF and/or a pointer to some docs? Thanks. Yours, Christian Holm Christensen ------------------------------------------- Address: Sankt Hansgade 23, 1. th. Phone: (+45) 35 35 96 91 DK-2200 Copenhagen N Cell: (+45) 28 82 16 23 Denmark Office: (+45) 353 25 305 Email: cholm@nbi.dk Web: www.nbi.dk/~cholm
This archive was generated by hypermail 2b30 : Tue Oct 16 2001 - 11:15:39 EDT