Hello We have all read all such sales words as "Performance: StreamBase processes messages at rates of up to hundreds of thousands of messages/second on a single CPUâa factor up to ten times faster performance vs. alternatives." But what is measured, incoming trades as events or outgoing aggregations as OHLC events or maybe just in goes out? What do you think?
streambase is a generic event processing engine, so it could be anything really.... their engine was developed for stuff besides financial applications. it's most applicable to tick data processing though, since bar data is much less voluminous and processing quickly is less difficult note that tradelink also processes ticks in hundreds of thousands per second, and unlike streambase it's totally free and open source... plus we focus exclusively on the financial industry unrelated business such as ecommerce and defense such as streambase. http://tradelink.googlecode.com
Hi tradelink So you think StreamBase Sales Speak could refer to anything, something like that was going through my mind. Well... my question was more in the like... CEP and not dedicated Trading Engines, btw. TradeLink only performs 50% of Esper, do you have a comment on that?
sure. the comparison you're making is not apples to apples, esper is a real time engine only.... so processing 500k events a second isn't actually very fast for an in memory realtime setup. I have seen people run in memory simulations with tradelink at 2-30million ticks per second. The 200-300k/sec historical performance in tradelink is based on historical simulations, something that's not possible to do with esper. Also note events and ticks aren't the same apple in terms of comparison either. A single tick could have multiple events/signals generated from it. so in the case of an in memory simulation above, tradelink events rates could be many multiples higher than the per tick number. The reason for this is that CEP frameworks are always going to add extra overhead in comparison to using if/then statements. Thus for the same if/then logic, tradelink should most always be faster than a CEP framework if they are both running on .net. If you choose to use CEP, you should make sure you're getting a return for this extra overhead in terms of saving you code or the time to write that code. TradeLink is geared for trading applications only, so we put tons of stuff in there to minimize the code you have to write. In a vanilla CEP framework you get none of this, plus you don't get broker or feed connections, you don't get a historical tick store and the ability to run historical simulations.... etc etc. It should be noted that there's no reason you can't use a framework like ESPER inside of tradelink. tradelink is fully .net compliant. http://tradelink.googlecode.com
Ohh I wasn't looking for a trading platform, just curious, because I have read a big amount of literature of the subject. The performance of Tradelink is however interesting, always nice to have some benchmarks while developing
Also they usually quote times for doing something very basic. It might take in one piece of information and just pass it on. That won't take very much time compared to doing heavy analysis.
why do you think you cannot run historical simulations through esper? or am i miss understanding something