"We’re using the cloud to plan more effectively," says Philip Jacob, senior director of risk management at Axioma, New York, who spoke at a CIO panel sponsored by Intralinks yesterday and then in a follow-up interview. The company is building a multi-asset-class risk management system for calculating potential losses on large books of complex assets. It provides stress testing under a variety of market scenarios…
The company is outsourcing those complex numerical calculations to the cloud rather than owning a massive farm of computers itself. "Because you’re using the cloud, the server farm is of an infinite size, you don’t have to make a decision up front on how much capacity you own," Jacob points out. "If you own your own data center, the speed at which you can run an analysis is driven by the number of CPUs in that farm. If you want to go faster, you have to make capex. Whereas if you do it on the cloud, you can on-demand call up more horsepower."
If Axioma owned all its own hardware, it might sit idle for 16 hours a day, Jacob notes.
Although Microsoft uses standard servers in its cloud rather than high-performance machines, Jacob says most risk calculations are easy to run in parallel across many machines, so that the desired speed can be achieved. "You need to describe your problem in a way that it can run well across lots of instances of commodity hardware," he notes.
As for the "s" word in cloud computing — security — Jacob says he covers it by carefully separating sensitive client data, which stays on the client’s premises within its firewall, from anonymized data that’s fed to the risk analytics engines in the cloud. "The only part we push out onto the cloud are the mathematical calculations," he says. "We get the benefit of the cloud without giving away any of the secret information."
Axioma gives its clients a small software program that calls out to Microsoft Azure on demand to perform risk analyses. "The normal workflow for enterprise risk is that after the markets close, there will be a lot of consolidation of market data that’s fed downstream to the risk management systems," Jacob explains. "Then before the markets open the following morning, the risk management function needs to understand what its exposures look like, so there’s an overnight window in which to generate all the results. As people start waking up, they will spot the risk concentrations and need to mitigate those risks. The cloud allows you to not limit the possible complexity in calculations. The cloud allows you to make everything efficient."
At many banks, Jacob points out, risk analysis is not a 24/7 activity, but rather a chore that causes workload spikes at intervals such as end of day, end of month and end of year, which is why leveraging cloud computing makes sense. So far, Axioma has launched some trial clients on a pure Azure framework.
One thing the cloud does, Jacob says, is let people stay in small groups but tackle large tasks. Collaboration requires people talking to each other, data and computation, he says. "The cloud can help solve two parts of the problem," says Jacob. "It can’t make people talk to each other, but it can provide a place where computations can be easily run, and where the results of those computations can be matched with output from another organization’s computations."