A Farewell to VPAC/V3 Alliance
Submitted by lev_lafayette on Mon, 12/21/2015 - 03:55
For the past eight years I've worked at the Victorian Partnership for Advanced Computing, also known as V3 Alliance, its trading name after merging with the Victorian eReserch Initiative. Today is my last official day, although I suspect I'll be doing "VPAC things" for a while yet.
When I came on-board, VPAC's main cluster was Edda, a 47-node Power5 system with 188 CPUs, running SuSE. Very shortly afterwards Tango came on board, a very impressive Opteron based system with 888 cores. Over the next five years it would 27 million CPU hours to more than 2 million jobs. Tango was the first HPC system that I worked on and, although it was significantly less powerful than its successor Sandy Bridge system Trifid (which, now with over 3000 cores, beat Tango's usage metrics in under a year), I still have fond very fond memories of that system, watching and helping it develop from initial installation to today, where but a single management node remains as part of the original system. After this of course, there was the numerous managed services which V3 took up as it became an organisation with national reach.
Provision of bare metal is one thing, but the best technical capacity is ineffective without the researcher involvement and participation to make use of the systems. This included at VPAC work with various collaboration tools such as Access Grid (the ideas now largely taken up by Google Hangouts and the like), TWiki, and many versions of Drupal. Perhaps a little less brilliant was the tragic rise and fall of ARCS, the Australian Research Collaboration Services, an experience which motivated me to complete an Masters in Business Administration (Technology Management). At least there I was taught that one should always seek to streamline your management overhead and expand your technical staff - not the other way around.
One issue which I took very seriously from the start was the need for technical documentation, and it needs to be reiterated now. Good documentation - like commenting in code - not only makes it clear to the authors what they are doing but also to future authors. On the equivalent management side there is the need for clear, transparent, and controlled policies for quality assurance. Having a structured change system in the form of project management is also very worthwhile. In all these cases the empirical evidence is overwhelming that whilst this may cost some extra initial time, but will certainly save time and improve productivity in the intermediate and longer term. In some ways this can be viewed like emergency contingencies, such as being a fire warden, first aid officer, or OH&S representative - all additional roles I had undertaken at VPAC.
After being fully re-integrated into VPAC after the ARCS experience I dedicated a lot more of my time to HPC training. From the middle of 2011 to now, well over one hundred days of classes were conducted to over fifteen hundred researchers from over twenty different institutions on Linux, batch job submissions, HPC architecture, sequential and parallel programming, and other related subjects. It is very pleasing that in its last months much of the material from these courses are being included in a range of publications that VPAC has allocated ISBNs for. Already such as luminaries such as David Beanland and John Gustafon have offered to write forwards to these publications.
As an organisation, VPAC/V3 Alliance can certainly be justly proud of its achievements over the past fifteen years and its skilled and often really quite brilliant staff. I cannot begin to describe how much I have learned by simply being in the same room as such people. Nor can I pretend that I am particularly happy with the fact that VPAC/V3 is closing down, or with how that closure is occurring. Whilst never an organisation that was particularly well-funded, VPAC/V3 nevertheless managed to produce two systems that were in the Top 500 supercomputer systems in the world at the time, as well as train and assist thousands of researchers and technical support staff in the field of high performance computing and parallel programming. The purpose of the organisation - it's esprit d'corp if you will - was to provide the effectiveness of researchers by the provision of computational facilities, by which they can improve the well-being of all.
Indeed, what does it benefit a person, even if they acquire great wealth and power, fails to make the world that we live in a better place?
For those who wish to remain in contact, I am now working have a new role at the University of Melbourne as an HPC Support and Training Office. My email there is xxxxxxxx at unimelb.edu.au ; my personal email is xxxxx at levlafayette.com
Dear Valued Colleagues,
The Victorian Partnership for Advanced Computing Ltd (VPAC) trading as V3 Alliance will cease its current business operations on December 31st, 2015 after 15 years of serving the Australian research community.
The Victorian universities that established VPAC in 2000 with financial support from the Victorian government have decided to dissolve this shared advanced computational platform as they develop alternate strategies to support their computational research needs.
During the past 15 years we have had the privilege to serve the Australian research community and be a significant part of assisting researchers across multi-disciplinary areas within academia, industry and government to develop new knowledge and create solutions that have benefited the greater public.
We thank you for your support over this period and wish you all well in your endeavours to create a better world founded on evidence based science.