Hi Denis,

Glad to hear that you are using the CEP as part of your application!

One possible solution to the problem you described is to review the EPAs
and the temporal context definitions and make sure the definitions are
defined taking into consideration performance and memory consumption.
Make sure that temporal contexts are terminated when no longer needed
(using event or relative time terminators). Otherwise the number of open
contexts can grow over time and cause the problem you mentioned.
In an EPA of type Sequence, the detection time grows significantly if more
and more input events are processed, a solution for that is to set the
participant event "Consumption" policy to "Consume".
Another good thing is to mark an EPA cardinality policy to "Single" if you
want it to be detected only once.

Another option is that there is internal CEP implementation problem. We
saw such behaviour in another CEP application and we currently analysing
it to see if there is an internal problem. In the other application, as a
temporal work around we restarted tomcat every month to clean the memory.

Since we are having many users and many support requests, we started to
document our answers in StackOverflow to help us and our users community.
It would be very helpful for us and the community if you could post this
question there. If you agree, please send me the link to your question,
and I will post this answer there as well. Thank you very much in
advanced!

(I cc fiware help list, in order to have this correspondence documented in
FIWARE issue system)

Best regards,
Tali

Tali Yatzkar Haham
Smart Decision Solutions
IBM Research - Haifa, Israel
tali@il.ibm.com 972-4-8296320

From: Denis Ermacora <ermacora@datamind.biz>
To: Tali Yatzkar-Haham/Haifa/IBM@IBMIL
Date: 21/08/2015 02:36 PM
Subject: Cep

Hi, Tali.
I wrote you some months ago about the Complex Event Processing sintax
and you gave me very useful information.
Now the application we developed that uses CEP as a Tomcat web service
is fully functional and works well.
The only problem I noticed is that the reaction time of the internal
engine in response to new events worsens rapidily. The more the engine
is fed with new events the higher the response time. I suppose it is due
to the fact that some of the internal events processed in previous calls
to the service remain in the engine internal tables. I wonder if there
is a way , for example, to make the CEP destroy explicitly all the
events within a given segmentation context. Thanks in advance for the
answer.

best regards,

Denis.

–
Denis Ermacora

Research and Development
DataMind S.r.l.
Parco Scientifico e Tecnologico
Via J. Linussio 51,
33100 Udine - Italy

Tel: 0432-629778
Fax: 0432-603887
ermacora@datamind.biz

C.F. & P.I. 02438900306

_______________________________________________
Fiware-tech-help mailing list
Fiware-tech-help@lists.fi-ware.org
https://lists.fi-ware.org/listinfo/fiware-tech-help

[Created via e-mail received from: Tali Yatzkar-Haham <TALI@il.ibm.com>]