WestGrid is getting a major boost thanks to a new $10 million partnership with IBM Canada, but a University of Alberta researcher
is warning a lack of support funding will make it difficult to fully capitalize on the investment.
A partnership of seven Western Canadian universities and research institutions founded in 1998, WestGrid provides researchers at member institutions with access to grid-enabled high performance computing and collaboration infrastructure.
After a competitive bidding process, WestGrid has selected IBM as a new partner. IBM has provided a package worth $10 million that includes deep discounts on 128 pSeries processors for the U of A, as well as software and technical support.
Jonathan Schaeffer, a computing science professor at the U of A and a WestGrid co-principal investigator, said WestGrid’s current infrastructure is all Silicon Graphics machines. IBM was anxious to break into the U of A and work with WestGrid, and gave them a generous discount on the new infrastructure.
“To be quite honest, there was absolutely no way we could afford the infrastructure we now have, it was only partnering with IBM and getting a very generous discount that we could do this,” said Schaeffer. “These machines outperform everything we currently have.”
Schaeffer said the new infrastructure brings capability computing to WestGrid, with much larger memory availability and lower communications costs between processors. The UBC facility, by comparison, is a capacity computing facility, with 1,700 conventional processors connected by a slower network with less memory.
“We now have capabilities above and beyond what you would see in a commodity computer, allowing people to solve problems you just couldn’t do on conventional machines,” said Schaeffer. “We have a lot of research projects that really need this kind of facility.”
One project that needs the IBM machines is Project Cyber Cell, which is trying to model a single cell. Schaeffer said a cell is incredibly complex, calling the ability to accurately model the life of a cell one of the Holy Grails of scientific research.
“The computation requires lots of CPUs, the CPUs must communicate quickly, it requires lots of memory, and it’s going to run for a long time,” said Schaeffer. “You can’t run this on a PC, or even a bunch of PCs, because you lose out by the lack of memory and high-speed communications.”
A team of 10 IT staff at the U of A will support the new infrastructure, as well as the existing SGI technology. They play an important role, making sure applications are ready to be run, and seeing if they can be tweaked to improve efficiency.
With WestGrid maintaining the current SGI infrastructure the IT team will now have two diverse sets of technology to support, but Schaeffer said they won’t be able to hire any more staff or dedicate more resources to the task, making it a real challenge.
He says it’s a missing link in Canada’s research and granting community. There’s a lot of money to buy shiny new machines, but not much money to support them. It’s a problem faced by institutions across Canada.
“When you’re buying a $10-million piece of equipment it’s a big investment, and you want to make sure you have enough people in place to get the most effective use out of that investment,” said Schaeffer. “That remains a huge challenge across Canada because there’s scant money available for it. It’s a real shame.”
Schaffer was part of a panel that mapped out a long-term plan for high performance computing in Canada, and he said the gap between money for infrastructure and support was identified as a major problem. They’ll be lobbying government to address the issue.
“Right now there’s nothing we can do. We’ve tapped all the sources we have to get money,” said Schaeffer.
WestGrid will get some long-distance tech support from IBM, as part of their in-kind and support package. Bernie Kollman, IBM Canada’s vice president, public sector Alberta, said that includes access to a technical advocate to exchange ideas on how to get the most from the infrastructure, and access to over 3,000 IBM scientists and engineers globally.
“The opportunity to collaborate with WestGrid, test ideas, and accelerate our own research and development was compelling to us,” said Kollman. “We’re very proud and pleased they selected IBM to be a partner with them.”
Integrated with WestGrid’s existing infrastructure, the new supercomputing facility includes two 64-processor P5 595 servers, each with 256 GB of DDR RAM memory and almost a Terabyte of local storage capacity, running IBM’s High Performance Computing software.
“It’s to give researchers the ability to test things they couldn’t previously test, and to share the infrastructure across the seven institutions of WestGrid,” said Kollman. “We see this as a good collaborative effort to support them in their research, and also to support education, which is a large part of what IBM focuses on.”
Comment: [email protected]
