|
|
||||
|
Bulletin 123 - 2004 Aug 25
1. HPCCC - fax, and Help Service telephone The HPCCC fax number, 03 9669 8112, is now working. The Help Service telephone is also working - 03 9669 8103 - we have set up our new 'phone system to allow multiple staff members to be able to answer the 'phone calls. 2. HPCCC - use of jumbo frame network There are several networks connecting to the SX-6/TX-7 system. One of the Gigabit/s networks runs jumbo frames - large data transferu units, which can lead to higher performance. The following are the names of jumbo-frame interfaces visible on the Bureau network side of the system. mistjf galejf mars1jf mars2jf fogjf cloudjf samsrv2jf tx7bmjf (The name tx7bmjf is a floating IP address which works whichever TX7s are available.) The jumbo-frame network is also available between the TX7s and cherax, and the following names are available: cherax-direct tx7-direct (The -direct names will provide the best connection we can provide at any time. tx7-direct should be used on cherax as this is the floating IP address.) For example, to copy data from gale to your home directory on the TX7s: rcp your_large_file_on_gale.dat tx7bmjf: To copy data from a TX7 to gale do: rcp your_large_file_on_tx7.dat galejf: and similarly for other systems. Please note that you will need correct entries in your destination host .rhosts file, e.g. on the TX7s, include the line galejf username Experiments to determine the speed-up indicated a 20 to 50% speed-up for a large file transfer - the results varied considerably with the load on the links, and file caching effects. The jumbo-frame links are the best available for large data transfers, and should be used solely for this purpose, not for things like interactive usage, rsh, etc, which may diminish the performance for others. 3. HPCCC - operational SX-6/TX7 jobs for CSIRO Would CSIRO users who have a continuing requirement for real-time capability from any of the systems at the HPCCC please notify HPCCC staff. We need to set up a minimum SX-6/TX7 system to maintain in the event of factors such as loss of power, and need to be able to easily identify all jobs that have to meet deadlines. We will then run those jobs in the 'rt' queues. 4. Removal of service on farrer (the portal) HPCbull 122.8 advised users to move from farrer to b2.hpsc.csiro.au. The CSIRO HPSC group needs to use farrer for a particular role, and so gives notice that we intend to remove farrer from general service on Wed 8th September - the names farrer and portal will then be re-directed to b2. If you are still using farrer (the portal), please move to b2. If there is software you need to continue to use which is not on b2, please let us know asap. 5. HPCCC SYSTEM CHANGES - NQS II and ERS II upgrades With the upgrades to NQS II and ERS ii installed on 11th August, (SYSTEM CHANGE NOTICE 2004-A006 NQS II & ERS II improved functionality & reporting) the qstat command now has another option to report accumulated processor usage, rather than reporting the processor time for a process in the job. The new option is -cl. The CSIRO IBM Xeon-based cluster systems at the HPCCC are now entering service - one ("Nelson") is for a project for CSIRO Marine Research and CSIRO Atmospheric Research. The other ("Burnet") is to be made available to any CSIRO research group, initially directed to groups that can make use of significant fractions of the system. Please form a queue. 7. CSIRO External Services Network (ESN) CSIRO is implementing a re-structure of its network and servers, to provide an External Services Network separated from its internal network. The ESN will provide external access via ssh and provide access to HPSC web services. The changes to bring this about will cause some changes to methods of accessing the HPCCC and CSIRO HPSC services from outside CSIRO.
External users, include users who wish to access HPCCC systems while travelling, will need to use public key based access. This is actually quite easy. We will send a CD with a key, plus instructions to people who register for external access. To indicate your need for external access please send a message to our area for handling problems with external access: 8. Altix (cherax) software upgrade - Propack The upgrade to Propack was finally completed on the morning of 12th August. Unfortunately, we did not anticipate the impacts on users and codes, particularly with changes to library locations and terminal interfaces. All codes should be at least re-linked. With the upgrades, you can now rcp files larger than 2 Gbyte to or from cherax. We are exploring better control of the batch system on cherax, now that large numbers of jobs have arrived. Since the Propack 3.0 upgrade on cherax, some man pages did not appear with the correct formatting, and the keycodes received from keyboards changed. This was due to a change to the standard system character encoding in the locale setting. Type "locale" to check. If it says LANG=en_AU.UTF-8 then log out and log in again. It should be set as export LANG=en_AU or export LANG=POSIX to cure the problems. We have now corrected the global default locale on cherax. 10. Altix (cherax) queue visibility Following a request from a user, we propose making the job queues on cherax visible to all users. Commands like qstat -a (the recommended command for seeing the jobs in the queues) will then show jobs for all users. Historically, there have been concerns about the privacy of user and job names. Users are reminded of the conditions of use which they have signed, particularly the confidentiality conditions. The change will be made on Wednesday 1st September.
|
|
Comments to: © Copyright 2010, CSIRO Australia Use of this web site and information available from it is subject to our Legal Notice and Disclaimer and Privacy Statement |