My name is John Easton and I work for IBM in the UK.  This use-case is not attributable to any specific institution (they are unwilling to have their name known), but is representative of a great many financial institutions who are all struggling with the same sorts of problems.

An automatic trading use case

A lot of financial markets organisations are trying to address the challenges they face around their day-to-day business.  In this space, the key thing is latency.  The faster you can return a response, the more chance you have of getting to the market first (beating your competitors).  So... based on some trigger (likely to be a financial datafeed) from which millions of messages will be arriving every second you need to:

- get the message
- understand the message (i.e. is this something I care about)

- if yes, then I need to invoke the appropriate calculations to make a trading decision.  I also need to perform all of the necessary calculations that allow me to comply with the regulatory framework I am operating within
- If no, then I just throw away the message

- upon completing the calculation, I need to drive the appropriate action to the trading system.

Most of the grids these organisations have in place today have 1000s (in some cases 10s of 1000s) of compute nodes.  These need to be serviced with relatively small (likely a few kB in this sort of scenario - but for extremely exotic options this could easily be GB) of data.

I need to turn these calculations around as fast as possible.  A simple calculation needs to complete in less than a millisecond.  A more involved calculation might take a few minutes, but this is calculation time - the grid infrastructure needs to do all of 'its thing' in a matter of a few milliseconds / microseconds.

All of the financial calculation code is typically legacy C++ and ideally needs to be used in this environment without requiring modification.


Hopefully this use-case is of interest to you?

Best wishes,
John
Senior Consulting IT Specialist and Technical Staff Member
IBM Systems & Technology Group Infrastructure Innovation
Int:  7-313796.  Ext:  +44-1256-343796

The IBM Academy is holding a conference on High Availability Best Practices from June 18-20 at the IBM Southbury Conference Center, Southbury CT, USA. Earlier the same week, a conference on Performance Engineering will be held at the same location. Paper submissions and attendees are encouraged. Details on http://w3.ibm.com/academy/ (under Active Calls for upcoming studies and conferences).



Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU