King's College London
Financial Mathematics
Centre for Financial Grid Computing
King's Logo

News October 2009

Nvidia supports Computational Finance at King's

The KCL Centre for Financial Grid Computing is grateful to Nvidia Corporation for their generous donation of Quadro GPUs for research into advanced GPU-based parallel computation, and to Professor Claudio Albanese for his work on this and related projects. The donation consists of eight Quadro 4800 GPU cards, each with 192 cores running at 1.2GHz. With continued software support from Apple the Centre can now explore applications to computational financial mathematics of both OpenCL and CUDA. See for example, the preprint including a normal quantile for GPUs available here.

GPU Computing on Apple Macs: How to get GPUs/CUDA/OpenCL working on Macs (being updated).It is now possible to run both OpenCL and CUDA applications on the ARTS cluster, and inded under Snow Leopard you can have both together. The image following, though non-financial, shows OpenCL and CUDA N-Body simulations running side by side under Snow Leopard under two Quadro boards.

About the Centre

The Financial Mathematics Group created the Centre for Financial Grid Computing in Spring 2008. The Centre for Financial Grid Computing at King's College London explores the use of advanced mathematical software in a grid environment. Apple, creator of the Macintosh and iPod, is helping to create the project's grid computing infrastructure through the Apple Research & Technology Support (ARTS) programme. Wolfram Research is supplying GridMathematica to the project. Equity fundamental and price data is being supplied by Thomson Reuters.

Compute status at October 2009: 64 Intel cores, 1536 CUDA/OpenCL cores.

Publicly Viewable Grid Resources   KCL Internal Grid Resources

Apple ARTS

The Apple Research and Technology Support programme (ARTS) represents a recognition of the importance of science in our society, an endorsement of the leading research institutions in Europe, and a celebration of the talent of researchers in all fields.

King's College London has joined CERN in Geneva and the Pasteur Institute in Paris on the list of ARTS institutions. ARTS Laureate Professor William Shaw has received an Apple solution to financial computing problems in the form of a cluster of 64-bit 8-core Mac Pro workstations. The system allows flexible deployment of parallel algorithms ranging from multiple 8-way computations to all cores being utilized in parallel. The system kicked off with a 48-core deployment of 2.8Ghz processors, with 36GB RAM, and now has further access to two XServe systems, so that 64-cores are available.

Our primary mathematical software sponsor is Wolfram Research, makers of GridMathematica.

In the King's model of financial grid computing, each core in the system is armed with a Mathematica kernel capable of doing advanced symbolic calculations requiring calculus and special functions, as well as traditional numerical processing. Previously, grids have been employed in a less mathematically sophisticated form, where each node does a purely numerical job. Also, significant program redefinition has been required to do efficient parallelization, which is an obstruction to industry deployment where the time of quantitative programmers is an expensive quantity. The new model allows both the parallelization of complex mathematics and minimal reprogramming.

   

Thomson Reuters for Academia

Thomson Reuters is the global information company providing indispensable information tailored for professionals in the financial services, media and corporate markets. Thomson Reuters information is trusted and drives decision making across the globe.

The King's Centre is currently working in partnership with Thomson Reuters, an agreement which has given us access to use the full range of data from Reuters Fundamentals and Reuters Estimates, for academic research on a number of our projects.

  Global Quantitative Research

Nomura's European quantitative research team supports the Centre for Financial Grid Computing and is working with the Centre on specific applications that will benefit from the high performance capabilities of a grid network.

Computational Financial Mathematics at King's College London

The research into financial grid computing is one of many threads of research carried out by the financial mathematics group. A major topic of interest is the mathematics and computation of dependency. The world's financial institutions are linked by a complex network of financial arrangements creating a highly dependent structure where an event in one part of the system can have profound consequences elsewhere. The risk position of one entity in the network can rarely be modelled in isolation, and more frequently requires extensive computation within models that take into account the relationship with many other entities in the network. This is one of many natural applications of grid computation. Other grid computation projects of interest include

Monte Carlo simulation for financial contracts

Risk calculations for large and complex portfolios

Solution of partial differential equations for contract valuation

Portfolio and basket optimization

Factor analysis and searching

The Financial Mathematics Group at King's College London hosts a weekly seminar attended by both academics and practitioners in the city of London, runs a large and successful MSc programme, and works with other colleges in the University of London running Computational Finance classes for PhD students within the London Graduate School in Mathematical Finance for PhD students.