Recent advancements in high performance computing technologies from SAS Institute Inc., including an in-memory analytics solution that uses Hewlett-Packard Co. hardware, are expanding possibilities for risk management.
HPC is changing risk management, according to Michael Stefanick, head of the risk management practice at SAS, by integrating both business functions and risk functions into the business decision process. “A business manager can actually look at the risk dynamically as he starts to make decisions,” he said.
Three high performance computing (HPC) solutions are available: in-database processing, distributed grid computing and in-memory. The in-memory analytics solution is the company’s most recent advancement and was formally announced this summer.
SAS says the technology enables simultaneous calculation of complex analytical calculations and instant access to results. The solutions will be generally available in late 2010, starting with risk management and then retail.
The HPC technology provides a way for financial institutions to attack large risk management problems in parallel, perform real time portfolio stress-testing and create detailed cash-flow projects, says SAS.
At a recent high-performance risk briefing for the financial industry in Toronto, Stefanick said applications of the technology include scenario analysis, covariance simulation, current exposure, cash flow analysis, conditional Value at Risk (VaR), Monte Carlo simulations, potential future exposure (PFE), liquidity and stress testing.
The distributed computing solution reduces the probability of loan default calculation time from 96 hours to four hours and the in-database solution reduces scoring of purchase behavior models from 4.5 hours to 60 seconds, he said. SAS in-memory analytics can reduce 8.8 billion portfolio VaR computations from 18 hours to less than three minutes, he said.
IT departments have struggled for years with the amount of historical they need to retain as well as the complex path-dependent mathematical calculations, said Stefanick in an interview with ComputerWorld Canada. And HPC is helping “not only to shorten that calculation time, but on the inputs in terms of the number of data elements that have to be stored,” he said.
“I can just take real time data from sources that are available, put it into this appliance, and it will sort out very quickly which ones are important and which ones aren’t,” said Stefanick. SAS’s HPC technology holds “all those results memory-resident, so the analyst only needs the ones that are important to them at the time of calculation,” he said.
High performance computing is changing risk management by bringing together the front office “earning money” team with the back office “risk control” team, said Nobel laureate economist Myron Scholes, a speaker at the SAS event.
A system that is both dynamic up front and systematized for the back end, he explained, moves it from being a reporting function to part of the actual operations of a company. “If you get answers quickly at low cost, then you bring it into part of your decision-making framework,” he said.
At the briefing, Scholes defined risk management as “the idea of managing under uncertainty” and suggested a change of focus. “We have to think in terms of risk all the time and embrace uncertainty,” he said.
“Risk management is always the idea that you are trying to measure risks … my view is that you are not really trying to measure risks, you are trying to understand how risk itself affects the decisions you make,” he said.
In-memory computing is opening up “a whole new world” for risk management, said Wes Gill, executive lead of risk management at SAS Canada.
Economics concepts such as liquidity and time horizons and shocks can now be modeled “in reasonable time frames, whereas before, people just weren’t doing it because the technology wasn’t there to do it,” he said.
Financial institutions can use the technology to improve their models and remove some of the inaccuracies, he said. “What we are able to do is get rid of some of the assumptions that we used to have to make to collapse data into aggregate numbers,” he said.
“This computation, because it runs so fast, allows us to basically go back and take a lot of those assumptions out in terms of aggregation, which makes the model a lot more accurate – and run the model through in a reasonable amount of time that management can use,” said Gill.
While the computation doesn’t replace management judgment, it allows management to “get the answer back in minutes as opposed to days,” he said.
“We can make it near real-time for them, so they can make a lot of decisions faster and they can make their models and the results from the models a lot more accurate,” said Gill. This also means doing “a lot models than they’ve ever been able to do in the past,” he said.
SAS’s in-memory computing solution operates on HP BladeSystem infrastructure with Linux. Gill said the “secret sauce” behind the technology is the ability to take “those millions or billions of calculations” and allocate them to over 2,000 plus CPUs at the same time.
A recent SAS white paper co-authored by Scholes and Stefanick, “Evolving from Quantitative Risk Management to a High-Performance Risk Management Analytic Framework,” suggests financial services institutions move to a single integrated and common risk, capital and management platform.
And SAS’s high-performance risk management appliance – which consists of an integrated risk and capital analytics mart, integrated risk analytics engine, integrated management analytics environment and integrated hardware environment that uses commodity hardware – provides this platform, according to the report.
“By having all of the core components already integrated on a common platform, less time and money can be spent on the infrastructure. Therefore, more time can be spent on analyzing the risk problems and optimizing the capital deployed,” states the report.
HPC is changing risk management, according to Stefanick, by integrating business functions and risk functions into the business decision process. “A business manager can actually look at the risk dynamically as he starts to make decisions,” he said.