Financial services

Interview:

The new frontier for actuarial modelling

2 March 2010

Hosted solutions for actuarial modelling are the answer to insurers’ IT needs, says Steve Foster at SunGard.

Insurers globally need to alleviate the IT burden of high performance computing (HPC) required to support increasingly complex actuarial modelling. Traditionally, demand for risk calculation processing is variable. It increases at peak times to support period-end reporting and it may also increase to support specific business conditions, such as new product development or fluctuations in the market. Most recently, insurers face the requirement for more transparency and accuracy in order to meet regulatory demands like Solvency II. This places a continued performance, capacity and optimisation load on a company’s IT infrastructure. As a result, insurers are looking for innovative ways to support these requirements while controlling IT costs. Finance on Windows spoke to Steve Foster, a senior manager at SunGard Consulting Services, to get his perspective.

You have recently participated in a project that brought together some of SunGard’s most valued assets and partners –the SunGard Infinity Platform, iWorks Prophet, and Microsoft in order to create a managed hosted environment for actuarial calculations. What was that like?
It has been a fantastic opportunity. iWorks Prophet Enterprise presents some interesting challenges, and I can see that a managed hosted service is a way to really help clients use the system more effectively. Having the right environment is key to this. It’s not just a question of mashing all the technology pieces together. You have to have an industrial-strength platform to deliver the service effectively.

HPC doesn’t just offer job scheduling; it also manages failures, reassigns workloads and prioritises tasks.

Steve Foster , SunGard SunGard Infinity and Microsoft HPC technologies are key pieces of the solution. You need resilience? SunGard Infinity does that. You want terabytes of replicated storage? SunGard Infinity does that too. You want to make best possible use of 1,000 processing cores for 30 solid days and a world-class grid, Microsoft Windows HPC Server takes care of that.

Why is it critical for insurers to look for hosted solutions for actuarial modelling?
I think there are two angles to this. Firstly, we are not talking about small amounts of hardware. Typically, we are talking hundreds of processing cores for a client. That’s a costly investment and ties up capital. It requires a lot of management and maintenance, and you have to worry about upgrade paths and whether your solution is best of breed. You also need to worry about disaster recovery and archiving massive amounts of data. A hosted solution takes away all that pain. There’s another angle around support for the application. A managed service can provide economies of scale and at the same time concentrate the expertise. It’s win-win.

Are there other factors compelling insurers to look for additional capacity provided by grid-enabled frameworks?
Most definitely. There are regulatory changes such as Solvency II, which are requiring insurers to value portfolios and measure risk more frequently. Many companies also recognise that more frequent assessments are useful internally for better management. All sides want more frequent and accurate reporting, which shortens the reporting cycle, and that means more processing power. Grid computing is the only viable solution.

What are some of the key benefits of this offering?
There are two things that I personally believe are key. First, there is no capital expenditure for the customer and that’s an important consideration when capital adequacy is important and the sums involved are non-trivial. Secondly, I think quality of service is important. Some organisations are moving from a situation where actuaries are running critical regulatory reports haphazardly on departmental desktops to a controlled, properly managed, audit-trailed system. The control is back where it should be.

What differentiates SunGard’s offering from everyone else?
SunGard already does a lot of hosted and managed hosted services for its products. My task was to bring that experience to iWorks Prophet and ensure the environment was right for the product and its client base. I never doubted that we’d succeed, but SunGard needed to appreciate the differences between an actuarial system and a payments or a trading system. We also used the iWorks Prophet User Group and iWorks Prophet Expert Group as a sounding board. Some of our customers already host iWorks Prophet Enterprise in house, so we also listened and learned a lot from their experiences. With their help, we figured out better ways to address disaster recovery, for example. We also helped our sister organisations like SunGard Infinity and Availability Services understand the range of different types of systems out there. They’d built superb facilities and platforms for high-demand transaction processing systems but the demands of systems like iWorks Prophet, which out of necessity are still batch-oriented, was something unfamiliar. In the end, we learned a lot through this process and found that some of our own assumptions had to change.

How is SunGard using Microsoft HPC to improve the efficiency of the iWorks Prophet environment?
When you have 1,000 cores spread over 50 servers and you still don’t know if you can meet your processing deadlines, you need to have something pretty special to manage your resources effectively. Microsoft Windows HPC Server 2008 provides a productive, cost-effective and high-performance computing solution that can be deployed, managed, and extended using familiar tools and technologies and enables broader adoption of HPC by providing a rich and integrated end-user experience scaling from the desktop application to the clusters. HPC doesn’t just offer job scheduling; it also manages failures, reassigns workloads and prioritises tasks. It is designed specifically for this environment, and we’re taking full advantage of its capabilities. iWorks Prophet output into Excel spreadsheets for reporting is a massive task. We believe that we will soon be able to exploit Microsoft HPC to cut the time spent on that task substantially.

What kind of performance and efficiency gains did you observe?
Impressive. Many factors come into play of course, but one client saw a job reduce from 33 hours to 33 minutes! The calculation engine and resource scheduling in iWorks Prophet Enterprise is better than in iWorks Prophet Professional and, just as important, you don’t need to spend valuable time working out the best run schedule, or restarting a job that failed just because a colleague switched off his or her PC. Enterprise grids can also be very large if required. Many of the approaches we have at the moment are for 500 or 600 cores top-end cores, where the client previously had 100 or so desktops of varying age and specification.

Looking ahead, what do you think is the next frontier for insurance technologies?
The end game will be scalable performance on demand. All our clients want it, and the reason we – or anybody else for that matter – can’t deliver is nothing to do with the technology. At certain times of the year, such as January, every single insurer on the planet needs maximum capacity. We can’t deliver capacity on demand if there is no spare capacity to give. What would help us deliver a better technology solution would be if the regulators staggered the reporting cycle for different companies, and we could then trade capacity between them. F

This article first appeared in the Spring 2010 edition of Finance on Windows magazine.

 

Add a comment