Adequation of grid computing to computation intensive operations in
the financial domain
The aim of this post-doctoral position of one year, is to contribute to
the definition of a software platform in order to run valuation and
portfolio risk analysis in the mathematical finance area. This platform
should provide the following features:
- good response time, thanks to the intensive usage of parallelism and distribution
- a security level compatible with the constraints raised by the financial institutions
- an easy to use platform, for the programmer and also for the end-user, which is not mandatorily a computer scientist
- a good level of portability through high-level middleware
The postdoctoral position is to be held within the OASIS
project, which is a joint research project between INRIA
Sophia-Antipolis, Univ. of Nice Sophia-Antipolis, and the CNRS I3S
laboratory, and in strong collaboration with the OMEGA
INRIA research team. OMEGA provides its competencies in mathematical
finance. OASIS develops ProActive, an environment for distributed
programming.
ProActive is a GRID Java
library (Source code under LGPL license) for parallel, distributed, and
concurrent computing, also featuring mobility and security in a
uniform framework. With a reduced set of simple primitives, ProActive
provides a comprehensive API to simplify the programming of Grid
Computing applications: distributed on Local Area Network (LAN), on
clusters of workstations, or on Internet Grids.
ProActive features the following:
* Asynchronous calls: Typed Messages
* Automatic future-based synchronizations: wait-by-necessity
* Migration, Mobile Agents
* Group Communications with dynamic group management
* Object Oriented SPMD
* Component-based programming and composition using Fractal
* Transparent, dynamic code loading (up and down)
* XML Deployment Descriptors
* File transfer mechanism over the Grid.
* Security
* Crossing Firewalls with SSH Tunneling
* A Peer-To-Peer infrastructure
* Fault Tolerance and Checkpointing mechanisms
The software platform for pricing and risk evaluation in finance will
be developed using ProActive, and will make usage of machines forming
Internet Grids or Global Computing Grids managed in a peer-to-peer
fashion (aka desktop grids).
In this research and development work, a particular attention to the
usage of the mechansims available in ProActive should be devoted. In
particular, the following requirements that the platform should fulfill
are as follows (listed from a high to a lower priority order):
- ability to tackle failures, while assuring a given degree of
performance, thanks to an adequate control of the various parallel and
communicating tasks (this requires the usage of the ProActive mobility
and group communication mechanisms; more globally, this calls for the
definition of adequate decomposition and scheduling patterns exhibiting
a high degree of parallelism geared towards highly heterogeneous and
volatile computing nodes that one may find on a grid)
- ability to make the software platform evolves, e.g in order to
easily enable the experimentation of various mathematical finance
algorithms (this calls for usage of the object-oriented and also,
component-oriented programming methodologies promoted by ProActive)
- ability to express and ensure security constraints, like
encryption, non repudation of the communications, authentication of the
objects that take part in the platform, taking into account their
location (this calls for the usage of the hierarchical and dynamic
distributed security mechanism of ProActive)
- high degree of portability, on any sort of grid middleware (use
of the high-level deployement mechanisms of ProActive that enable to
provision then launching of distributed computations on any kind of
computation infrastructure).
A first version for
such a platform using ProActive already exists. The aim is to pursue
its design and development. In particular, the postdoc should focus
his/her research on the definition of adequate programming patterns
(generic, adaptative, fault-resilient, highly performant) devoted to
parallelism exploitation on computing grids for pricing of financial
options, risk analysis. For the time being, only the ProActive
programming for European options pricing, with and without barriers has
been addressed. The postdoc should also devote time to Value at Risk
calculation models, and hedge funds specificities.
In any case, we are interested by non trivial scenario, due to the
large size of trading portfolios, the heterogeneiry of their content
(various types of assets and options, like European, American, or Asian
ones), so that pricing and valuations may require huge amount of
computing power. We aim at reducing the associated computation time (a
few seconds instead of a few minutes in the front offices of
stock markets, or a few hours instead of a few days in the case of risk
evaluation at the global level of the financial institution)
We are expecting a young researcher skilled in distributed Java
programming, while also already aware of the principles and
computational methods required in mathematical finance (e.g such as
Monte Carlo simulation techniques), and possibly also, skilled in the
parallelisation of numerical methods.