A new era of supercomputing
Andrew Carr looks at an on-demand approach to high-performance processing and finds benefits for the channel
Go back 40 years or so, to the beginning of the supercomputing era, and technology had limited application beyond weather forecasts and telephone switching. Typically, individual applications cost lots of money, expertise, and other resources. High-performance computing (HPC) remained narrowly focused and dominated by large supercomputing centres, research groups and academic institutions.
Only these kinds of organisations were generally able to attract European funding for specific programmes or activities, or obtain access to large IT budgets for one-off projects without having to conform to a cost-focused business case. Typically, they would buy complicated IT systems that would incur significant implementation and maintenance costs and need regular technology refreshes.
Over time, we have seen HPC develop to meet an ever-expanding portfolio of needs and challenges, and the advent of cloud is changing the whole game, delivering new ways to produce performance computing, from dynamic provisioning to automation and orchestration. It is also delivering new ways to consume IT - as a service, on demand, or via configurable models.
Instead of aspiring to be the smartest buyers of HPC, firms increasingly want to be the smartest users of it.
Cloud is helping democratise HPC within all industrial sectors by lowering the barriers to entry, cutting costs by offering the opportunity to pool resources, shifting expenditure to the opex column, and giving users the ability to share workloads within communities or even industry sectors.
Cloud HPC, sometimes dubbed HPC on demand, has the potential to appeal to any organisation with fluctuating power needs and a limited budget. Usually it involves the solutions supplier investing in infrastructure that gives prospective customers the opportunity to buy access to that computing resource, rather than having to make an upfront investment in a complex IT hardware implementation.
Such a new service is likely to attract attention from large enterprises that already have their own computing resources but want additional processing power to deal with workload peaks. It may also appeal to SMBs and design offices that only have basic computing resources and would like to seize the opportunity to outsource hardware and technical support. This will help SMBs and design offices concentrate on their core businesses.
But the greatest benefit, especially in an economic downturn, is the potential that HPC on demand offers to kick-start economic recovery. It could do this by helping the long-beleaguered manufacturing and engineering sectors achieve operational efficiency.
Cloud HPC can drive “time to insight”: the time taken between the presentation of a problem and figuring out how to solve it. Manufacturers can shorten their design phase, add time to the build stage, and reduce their overall time to market.
Many engineering companies struggle to accurately predict their future workload. In most cases, there will be no predictable pattern of need. If a company implements HPC in a traditional way, it runs the risk of having to overprovision or being limited in the design work it can do, because it may not have enough of the right kit.
On-demand HPC offers such companies the agility to scale up when required. They would no longer need to worry about the complexities of running their own environment or the negative effect this might have on their profitability. Instead, they could tap into the available computing capability across the web as and when needed.
They have less need for physical infrastructures and lower maintenance requirements, while still having access to high-performance tools for innovation. The developers of supercomputers back in the 1960s would have been amazed and delighted by the possibilities - and providers of IT to businesses should think about it too.
Andrew Carr is UK and Ireland sales director at Bull Information Systems