Compute Canada is returning to its user community to help map out the future research needs for advanced research computing (HPC) between 2017 and 2020. It is asking for bold world-class proposals to shape its request for new funding in the next round of the Canada Foundation for Innovation's Major Facilities Infrastructure competition. The consultation comes at a critical time for Canada's diversifying base HPC users who are increasingly coming from non-traditional disciplines and typically gravitate to where the best facilities and computing power reside.

"People tend to work where the infrastructure is and we need to have the tools in Canada to allow them to thrive," says Dr Dugan O'Neil, Compute Canada's chief science officer. "This is a starting point to create a common infrastructure and platform … For a lot of scientists, they can't compete without the basic advanced research computing being in place."

Dubbed SPARC2, the consultation follows a similar process to the first Sustainable Planning for Advanced Research Computing (SPARC) consultation last year. That produced more than 20 papers which informed CC's planning and assisted in its successful mid-term review and request for CFI funding, resulting in a crucial $30-million award .

While CFI has been instrumental in supporting the HPC research community, its last cash infusion took place when the state of HPC infrastructure was sliding into obsolescence.

"Before the $30 million from the CFI, we were on a path where most systems had to be decommissioned. We're now replacing everything we were turning off," says O'Neil. "It was a great stabilizing step but it doesn't push into a new area of resource provision. There's a global race for competing power and we need to help researchers stay competitive."

CFI was awarded $100 million in the last Budget for digital research infrastructure and Compute Canada officials are waiting for details on how the agency plans to allocate the new funding (R$, April 30/15).

SPARC2 is intended to solicit input for the future needs of ARC providers and users. Submitted papers are requested to provide information on:

• the area science to be studied

• types of ARC currently being used (storage, compute, cloud, portals and gateways)

• future growth projections

• data needs

• types of computation to be used (serial or parallel)

• software/middleware requirements

• networking (multiple sites).

"The speed of change in ARC is rapid. The science projects we serve have changed and put a strain on the system. Groups are forming and ready to talk," says O'Neil, adding that the impact of the original SPARC funding has yet to be felt as new, more powerful systems are coming on stream more slowly than those being decommissioned. "In this round we want to know what (users) need to be competitive in the world."

Diversifying user base

Adding to the complexity of determining Compute Canada's future needs is the diversification of the research base requiring ARC. Dugan says the ascent of cloud computing and massive parallelization have driven demand for accelerators — graphic interface devices most often used in conjunction with the CPUs to accelerate computations that require a large amount of algebraic operations.

"We're adding non-traditional disciplines and the way they want to use ARC equipment is different," says O'Neil. "We're now serving people who weren't there five or 10 years ago."

Canada isn't the only nation focussed on boosting cloud-based systems, data sharing services and specialized software. Demand is also up for very large systems — an area where Canada currently falls short of the leading nations.

"Some of our researchers go international for computational resources but it's often hard to get foreign access," says O'Neil. "Round I (of SPARC) gave people hope and Round II is to take a step forward. The world won't stand still."

R$