Bulletin 153 - 2006 March 17

1. SX-6/TX7 Systems interruptions
2. Memory scheduling
3. Timezone changes
4. Change of default versions of netCDF libraries and nco utilities
5. Closing Old REQs
6. ssh without passwords, and security
7. HPCCC SX compiler policy
8. Brochure on CSIRO HPSC
9. Parallel Coupled Models Workshop
10. cherax downtime

During the TX7 software upgrade on the morning of 6th March, a problem occurred, and both TX7s went down. The event is being investigated by NEC. The HPCCC apologises for the unanticipated interruption to service.

[ page top ]

A new version of qsubnew was recently installed across all the NQSII client platforms.

This version uses the memory requested by a job as one of the factors to be considered for deciding where to place jobs. It also reduces the weight given to URGENT jobs over SPECIAL jobs, and SPECIAL jobs over NORMAL jobs.

These changes should reduce the number of high-priority jobs unnecessarily displacing large-memory jobs, and improve overall system throughput, as well as continuing to provide expedited service for operational work.

Please test qsubnew and report any difficulties as soon as possible. If no problems are reported, the HPCCC plans to replace the qsub wrapper script with this new version on Tuesday 28th March.

Change notice 2006-A006 gives further information.

[ page top ]

Daylight saving time is extended this year owing to the Commonwealth Games.

Many systems have automated daylight-standard-time correction features in the operating systems, keyed to known dates. This one-time delay in reverting to standard time is not accounted for in the automated features. Manually setting the standard time zone delay has been found to have consequences on certain systems, and significant administration work, so special corrections are not being done.

Please note that as a result, many of our systems will be 1 hour in error in local time from 26th March to 2nd April.

[ page top ]

At present, old versions of netCDF libraries and nco utilities are the defaults on the HPCCC systems.

On Tuesday 28th March, we plan to make the latest versions the defaults on the SX-6s, the TX7s, gale, cherax and farrer.

The latest versions will be 3.0.1 or variants for nco utilities (except on farrer), and 3.6.0 or variants for netCDF. Old versions will continue to be available.

The cherax version of nco now has the OPeNDAP support enabled [req #6438].

Change notice 2006-A007 gives further information.

[ page top ]

We have found it necessary to remove those REQs that have been inactive to better manage new and active REQs. As part of this process Phil Tannenbaum has been reviewing and "Resolving" REQs that are both very old and in various zombie states.

• questions that have been answered or discussed, and have not had any interaction in 12+ months
• PSRs that NEC has since claimed to have corrected in previous or near-future releases (from SX-5 and very early SX-6 days)
• phantom problems that have occurred once and are essentially not reproducible or at great effort to investigate further
• problems on old system configurations where ongoing improvements have been made
• problems on systems or software that have since been replaced

In each case there has been no action by either the user or the HPCCC in over 12 months.

To minimise overhead, and because of the age the REQs have been dormant, Phil has often not sent email notifications to the submitting users; however, in each case the project/user/problem is known by the HPCCC to have progressed since the respective REQ was last acted on.

These "Resolved" REQs can be reviewed by checking the "Resolved" category, sorting by "Age" and checking those Phil has "Resolved". Each has a closing note as to why it was set to "Resolved." This will be an ongoing process targeting a REQ system where each open REQ will be topical, current, and better managed to resolution.

Users with "Resolved" issues that are still considered important should contact

As REQs get within a 12 month currency window of being actively worked on, emails will be sent to submitters prior to any action to "resolve" them being taken.

[ page top ]

It is possible to set up ssh using private and public keys so that logins and file transfers between machines can be accomplished without the need to enter a password every time, i.e. set up a web of trust between two machines.

Please please consult the ssh-without-passwds how-to guide at

- this covers both ssh1 and ssh2.

It is important that users do not store un-encrypted passwords anywhere on the systems.

commands where available, to avoid storing passwords.

[ page top ]

The HPCCC desires to have a stable SX-6/TX7 system during 2006, with minimal disruptions and avoiding unnecessary software upgrades.

The topic of SX compilers was discussed at the recent HPCCC Users' Liaison Committee Meeting, with various options being presented, to assure that a known stable compiler was always the default, but to provide users with access to the latest compiler that has been vetted for possible use.

Representatives agreed with the following plan:

1. The R313 compiler will remain the "default" SX FORTRAN compiler until a substantially improved compiler is available that has significant benefits for users, or roughly 1 year of accumulated corrections have been received.
2. The "latest" compiler will be the most recent provided by NEC that has passed HPCCC tests, i.e. performance is minimally no worse than the previous compiler version.
3. The "latest" compiler might change multiple times prior to the default compiler be upgraded.
4. New compilers agreed to have substantial benefits or corrections will be fast tracked through user testing and into "default" status.
5. The "default" compiler will otherwise be upgraded roughly annually to include cumulative corrections.

Note that the default compiler on cherax and farrer is still version 302, although version 313 is available.

[ page top ]

A brochure entitled 'eScience: delivering diversity' is available from http://intra.hpsc.csiro.au/ .

The brochure shows some of the applications using high performance computing in CSIRO.

[ page top ]

The APAC Computational Tools and Techniques project is running an informal one-day workshop on parallel coupled models at ANU on April 10th.

Many interesting applications in computational science are built around coupled models where separate models for different parts of a system are linked together to model the whole system. These individual parts may describe very different physics, use varying scales of measurement and exist as stand-alone modelling codes that use different parallel regimes.

For further details and expressions of interest please visit

For enquiries and registration, please email

[ page top ]

cherax will be out of service on the morning of Saturday 1st April, to enable instrumentation in the fibre channel switch, to allow analysis of i/o usage.

[ page top ]