|
|
||||
|
Bulletin 187 - 2009 March 26
The Bureau of Meteorology and ANU are pleased to announce that contracts have been signed for the acquisition of new supercomputer systems from Sun Microsystems. The announcement can be read at http://www.bom.gov.au/announcements/media_releases/ho/20090319.shtml More details of the systems will be available shortly. A Sun seminar on the Bureau's supercomputer system for high performance computing application developers is planned to be held during the week of 20th-24th April. [ page top ] 2. CSIRO - New NCI National Facility Cluster available The following announcement was recently made to all NCI National Facility users. A new production compute service is now available for use by all NCI National Facility users. It is a 156 compute node (1248 cpu) SGI XE Cluster (http://www.sgi.com/products/servers/altix/xe/). This system provides sustained performance equal to about 60% of the SGI AC Altix Cluster so it is a considerable addition to the Facility's computational capacity. The per cpu performance of the new system compared to AC will vary from application to application but on average is quite similar. Since the charging rates for the two systems are the same, we encourage all users to make the comparison for themselves to decide which system provides them with the most compute results for their project allocation. You can expect to see up to a factor of two difference in performance in favour of either system depending on your application. The new system is also being used to trial and develop system software for the next major National Facility system. In general, the user visible components of the system software should be familiar to you - the Linux operating system, the filesystem structure and policies, the Intel compilers, the queueing system and queues, etc. However a very different underlying system configuration is being trialed and teething problems are to be expected. Bear with us while we try to develop a robust, scalable system for the future. The login node of the new SGI XE cluster is xe.nci.org.au - all NF users attached to an active compute project should have an account on that system with the same username and password as AC. Details of the XE system and its use can be found at http://nf.nci.org.au/facilities/http://nf.nci.org.au/facilities/userguide/ In short, the system currently has 156 compute nodes each with two 64bit Xeon (x86_64) quad-core processors, i.e. each node effectively has 8 cpus. Most nodes have 2GB of memory per cpu, a small number have 4GB per cpu. CSIRO access to NCI facilities at the Australian National University can be obtained through either the CSIRO partner share or the NCI Merit Allocation Scheme (MAS). Applications for CSIRO Partner time are approved by Rob Bell. This application can be accessed at https://nf.nci.org.au/accounts/projects_new/partner_user.php For more information about the NCI Merit Allocation Scheme please read https://nf.nci.org.au/accounts/projects_new/MAC.php [ page top ] 3. Traffic light upgrade The traffic light system at http://www.hpccc.gov.au/hpccc/system_status/ and http://intra.hpsc.csiro.au/traffic/servicestatus.xml has been upgraded recently to provide graphs of software license usage in the last 24 hours. This information is available through the 'more' button, which also gives information about the number of licences available, and current usage. [ page top ] 4. Data Areas on cherax We have commenced moving significant shared data holdings out of users' home directories on the CSIRO Data Store, to enhance the access and management. On Tuesday 31st March at 10:00, the IPCC files in the area ~ofa001/IPCC will be moved to a new area. The new area will be accessed through the variables $IPCC and $CMIP3, which will be directory names, and will be set by the command: module load IPCC-CMIP3 Further areas can be set up as users and groups desire. Please provide answers to the following:
[ page top ] 5. New CSIRO ASC Software
[ page top ] 6. CSIRO ASC Software Upgrades The following have been recently installed:
[ page top ] 7. burnet scheduler The burnet scheduler (moab) has been updated and now supports requesting vmem in round numbers. For example, it will be possible to specify vmem=1GB rather than vmem=1000MB Note: Jobs specifying vmem=4GB are still problematic if they go to 32-bit nodes as setting a limit >=4GB fails on a 32-bit system. vmem=4095MB is OK though. It is recommended that you request close to what is actually needed. A summary of vmem usage is printed at the end of each job. [ page top ]
|
|
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 |