Why would a trading firm that specializes in high frequency trading use a product like TIBCO? Imo, this middleware will only slow you down and you are better off witting your own middleware. Maybe TIBCO does more than I think it does, or maybe performance has recently improved...
tibco was the standard middleware years ago (imo, mainly because reuters) and now firms have it only because of legacy purposes. i dont know of any firm starting with tibco now; i do know several that recently put in 29west. but if your going for absolute speed (which most firms don't need) then an inhouse solution seems to be the way to go.
Thanks for the response. That makes sense about TIBCO. As far as 29West, I don't get it. All that does is give you "reliable multicast" which can easily be implemented by adding a tiny header with sequence numbers. Maybe I am missing something...
One of the features of TIBCO that everyone liked was the ability to run redundant processing over their middleware; ie TIBCO could run a 2nd system in virtual lockstep which is great for "software fault tolerance" and failover. That's pretty much the only reason I ever heard from anyone with TIBCO deployed.
Nitro, I agree. Today I can't think of any reason to run Tibco. 5-8 years ago they were pretty much the only "low latency" game in town. Ultimately, the type of failover and redundancy that Tibco provided were often easily added in the application domain.
We've got a TIBCO setup in our framework and it's contributing a 2 millisecond lag that we need to kill. It's awful and we want to get rid of it so bad, but the reason we have it is because of some vendor product that uses it. But yes, stay away from this garbage. Roll your own.
Yes, and ActiveMQ and...a bunch of other MQs that are free and open source these days. If you don't see the benefit of messaging infrastructure you probably don't need it. Sure, you can roll your own clustering, failover etc. and use UDP multi-casting instead of MQ topics. Indeed, it may not be relevant at all for your trading infrastructure especially if you have certain non-functional requirements that can't be met by the MoM. With technology built on MQs such as Apache ServiceMix (http://servicemix.apache.org/) or Mule (http://www.mulesource.org/display/MULE/Home) server-side integration and aggregation becomes like using Lego. Again, none of that may be relevant to YOUR trading infrastructure which is why you might not see any benefit to it.
I just claim that mixing high frequency trading and this stuff can be hazardous to your pocket book. Of course is has value in general.