|
Algorithmic Trading
< Previous Next >
Re: High frequencey trading on Windows and .NET
Tanveer Ansari / Crystal
5 Jul 2008 6:19PM ET STAC has benchmarked some commercial feed handlers for market data such as WOMBAT.
Getting more detailed as to what part of the processing needs this performance -
For the processing of the raw market data feeds I am leaning towards buying a commercial product instead of writing each feed handler by hand for each feed. Some commercial feed handlers have an advertised latency of sub millisecond from the time a feed(say SIAC OPRA feed) hits their system to when the data is available to a client application. Market connectivity, network speed and hardware can all be best in class.
That leaves the business logic and this is what needs the low latency. The algo/logic itself is simple but this logic needs to run for several thousand different items, each potentially having multiple updates per second. The required peak messages rate) each of which will cause the business logic algo to be triggered is about 500,000 messages per second
> STAC Research are trying to set up some benchmarks around securities
> technology generally, currently they have a couple around market data
> but are also looking at the transactional message part. They are a for
> profit company and not affiliated with FPL, see
> http://www.stacresearch.com/
>
> Cheers Kevin Houstoun
>
> > Exactly, and if you're after ms probably many other parameters. Does
> > any one know of a "benchmark"?
> > > I am a bit surprised with everyone throwing numbers on the table
> > > without detailing much their meaning?
> > >
> > > - 100ms to do what? tp take a message from the market and process
> > > it? What type of message? what type of processing?
> > > - Should the hardware not be taken in consideration when it comes to
> > > discussing performances?
> > > - What about Network speed? Market Connectivity?
> > >
> > > Regis
> > >
Re: High frequencey trading on Windows and .NET
Tanveer Ansari / Crystal 5 Jul 2008 6:19PM ET
|
@FIXProtocol
