Monday, April 26, 2010

Letter to the editor in the Communications of ACM

Paul Vincent attracted our attention to an article in the Communication of the ACM written by Julian Hyde from SQLstream, and wondered about some of the assertions that were expressed in this article. Reading the original article, I indeed found some inaccurate assertions that somehow missed in the review process of the CACM magazine. After communicating with the editor-in-chief of CACM, he suggested that I'll write a short response (and the author will have a right to respond to the response). It takes some time, but today I got the electronic copy of CACM and found that my response, and the response of Mr. Hyde to my response have been published. You can read for yourself (under the title - "event processing anywhere"). In essence I have mentioned three inaccuracies in the original article:

1. On whether event processing is a blanket term for stream query systems: actually event processing is much broader term then stream query
2. There is a religious war between SQL and non-SQL vendor. I have not noticed the religious war, many vendors have both SQL and non-SQL interfaces.
3. Event processing is restricted to one application type in the financial services area, while other application areas are neglected; this is indeed a common misconception, and I have written about it in the past.

Mr. Hyde took his right of response by saying that the event processing vendors fail to satisfy the expectations of the BI community, this is a new claim, which does not seem to respond to any of the issues above.

Paul Vincent in his Blog's post raises another interesting assertion from the CACM article that streaming query engine is a new technology -- well, new is a relative term.

Bottom line: while people are used to view marketing material with a grain of salt, it is typically expected from a respectable magazine which employs review processes to be more accurate, since people tend to take it as a reliable source.


Hans said...

Rather than a religious war, every vendor implies (in blogs) or just says (in sales meetings) that the technology of the others is "interesting, but ultimately misguided".

This definitely happens, I have noticed it.

Opher Etzion said...

Hi Hans.

Of course vendors are trying to emphasize their perceived strength, but there is not really a religious war here like we have seen in the past in some communities around language styles. Moreover, the bigger vendors like - IBM, Oracle, TIBCO support multiple programming styles for various scenarios.



Rainer von Ammon said...

hmmm, I'm not so sure reagrding the right EPL approach discussion, Opher. Actually Oracle has only what was acquired with BEA event server and the CQL of Oracle, because BEA used Esper as SQL approach. Tibco does not use SQL-like approach, Apama has its own language as well, Streambase is SQL-like, they all have additionally a "studio" as a graphical modelling which facilitates a quicker start for newbies, but experts will work with the EPL. Only IBM has a lot of different appraches because IBM acquired AptSoft, etc., had always its AMiT, right? And when I remember the forum contributions of David Luckham's complexevents and Tim Bass's cyberstrategics, there was a bit a discussion about the appropriate EPL approach in the past years. But I would say the most have agrred with that you cannot use SQL-like approach for AlgoTrading, too much programming, or Starview would say SQL-like would not work in their Semiconductor application area.

But what you told him is right

Opher Etzion said...

Hi Rainer -- IBM has Infosphere Streams which has a stream based language (not SQL), and Websphere Business Event (ECA rules); Oracle has the CQL language, but also active database with rules, TIBCO has its Tibco Business Events which is based on RETE rules, but according to Paul Vincent's Blog it also has a stream processing capabilities. There are various algorithmic trading platforms that are based on SQL oriented languages thus those who Streambase, and some that are based on other languages - like the Apama stream language, so we have a big variety. I think that one of the discussion topics in Dagstuhl, when we discuss standards is --- what the future of event processing languages should be.



Rainer von Ammon said...

Hi Opher,
yes, let's see what we will discuss at Dagstuhl seminar, will be a great event.

Regarding SQL-like EPL + AlgoTrading: I would be careful, so far it was told me there is a little bit SQL-like code, but the rest is written in Java


Paul Vincent said...

Hi Opher, Rainer - yes, TIBCO BusinessEvents has (since version 3) supported both Rete rules (if-then production rules) as well as continuous queries (select-from-policy SQL-like).

I've not heard any religious debates of the type "lets program this in rules vs no lets use queries".

Both these (and other) paradigms have their place.

PS: good work Opher on the response.