Derek Furguson of Bear Stearns (now JPMorgan Chase) has a good article in .NET Developer Journal about how to apply genetic algorithms and grid computing to the problem of market timing in stock trading. I was pleased to see that he chose to implement his algorithms using the Digipede Network.
His article is in two parts, and this first part provides a good overview of the complex problem he’s facing — he confronts issues in financial modeling, data sources, genetic models and grid computing. As a result, Part One does not dig too deeply into coding details. But it’s worth a read — you’ll understand the architectural decisions he’s facing, and how he’s planning to address them. Plus, from what I’ve heard about Part Two (which will be out in June), there’s plenty of detail (and code) coming.
This is the second time in two months that we’ve seen influential financial modelers implement their public examples using the Digipede Network (see also Matt Davey’s recent Dr. Dobb’s article).
This is consistent with what we’re seeing from customers. While there are many grid offerings in the market, there seems to be a growing consensus that if you use .NET, there are significant advantages to working with a grid solution built on .NET. Or conversely, there’s no point trying to fit a square peg into a round hole — i.e., there’s no point trying to graft a .NET application onto a grid built for other technologies when a better option exists.
This is the “application centric” view — grids should follow applications, making it easier for developers to adapt applications to a grid, even if that means limiting the options for running those applications to a particular set of resources (in Digipede’s case, Windows machines running .NET).
The other view is “infrastructure centric” — that OS should not matter, that a grid should allow applications to be deployed across all resources, even if that means restricting the application technologies and development patterns allowed for such deployment.
Digipede has been unapologetically in the “application centric” camp for five years now, but what do others think? Has Derek made a wise choice by trading off ease of development for deployment limited to a single OS? We think so, but let’s hear from you!