Zürcher Kantonalbank accelerates risk assessment with an IBM grid

Zürcher Kantonalbank (ZKB) in Switzerland provides private, retail and commercial banking services for both individuals and corporations. With around $100 billion of customer assets under management, a balance sheet of over $7 billion and more than 4,200 staff in 109 branches in 2005, ZKB is a growing force in the investment management market, returning annual profits of around $600 million.

Companies in Switzerland operate with a high degree of interdependence, and it is important for ZKB, which provides credit for these companies, to understand how the fortunes of one can affect the others—for example, whether smaller suppliers would be able to handle the bankruptcy of a larger company. ZKB knew that it would be possible to use advanced computer modeling algorithms to gain a clearer view of its financial position, reducing the amount of capital it would need to offset against potential bad debt and increasing profitability.

Without such a modeling system, the bank had to rely on regulatory credit risk figures, which did not take into account factors like diversification, structure of credits, repayments and amortization, and were consequently inflated. As credit represents one of the bank’s main income streams, the need to find an IT solution capable of running complex financial models at high speed was becoming critical for the business.


Opening a path to performance

ZKB considered solutions from several vendors before deciding to implement a grid solution designed by IBM Global Technology Services.

“The pre-sales service from IBM won us over,” explains Dr Daniel Egloff, Head of Financial Computing at ZKB. “The IBM consultant listened to our issues and offered to perform a proof-of-concept at the IBM Grid Design Centre in Montpellier, France. The ability to determine our exact requirements before implementation was a crucial factor in our decision—and it convinced us that the solution would work.”

The proof-of-concept was vital because ZKB did not consider any of the commercially available risk applications to be suited to its needs. The bank’s ambitious strategy was to design its own state-of-the-art system based entirely on open source technology and running on a GNU/Debian Linux operating system.

“We wanted to prove that open source is a possibility for the banking services industry,” explains Dr Egloff. “And of course, the choice of Linux and other open source software kept our licensing costs at a minimum.”

Banking on the skills of its in-house team, ZKB developed its own financial modeling application to calculate credit risk. The team also built a scheduling and object distribution system based on MP, and use the Sun Grid Engine to provide a front-end queuing system.

The grid consists of several IBM Cluster 1350 servers with 45 IBM System x model 336 servers as computational nodes. The extreme scalability of the grid architecture will make it simple to add more machines in future to increase processing power—in fact, ZKB has already tripled the number of calculation nodes since the start of the project.

“The main features of the IBM solution are scalability, flexibility and performance,” says Dr Egloff. “We can easily add nodes to the network and divide them between production and test and development clusters to balance the workload dynamically. And with the grid architecture, we exploit around 90 percent of the available processing capacity—as opposed to an estimated 40 percent for alternative solutions.”

“Efficient use of hardware resources is important because our key requirement is computing power. Our financial modeling calculations involve up to 10TB of data, so a single calculation can take anything from 60 to 100 hours, even with the new solution. Without the grid technology from IBM, combined with their High Performance Computing and Linux cluster know-how, performing such calculations within a useful timeframe would simply not be possible.”


Cutting capital requirements

The grid is simple to maintain and administer, reducing workload on IT staff and freeing them to provide support and development services to the rest of the company. There are plans to extend access to the grid so that other departments within the bank can benefit from its computing power—for example, SunGard FRONT ARENA will soon be deployed on the grid to provide fast pricing of structured products.

“We would like to extend the grid’s financial modeling potential too, so that we can run all kinds of Monte Carlo simulations,” adds Dr Egloff. “It may even be possible to use the grid to provide services to other banks, potentially increasing revenues.”

The low cost of implementation—using open source software and extending the value of existing hardware investments—combined with the huge improvement in the accuracy of credit risk figures meant that ZKB saw a full return on investment after only the second computational run.

“Because we no longer have to use the regulatory risk figures—which are approximately twice as high as our own—we can dramatically reduce capital costs, making our credit operation far more profitable and at the same time more competitive in a market with shrinking margins,” concludes Dr Egloff. “The IBM grid computing solution has given us the power we need to calculate credit risk accurately, and its flexible architecture will enable us to embrace more complex financial models in the future without increasing costs.”