Hi Suho, 

Thanks for your answer. Currently, I haven't any implementation or
performance numbers.

Our proposal is independent on the partition of rule structure itself. Thus,
it will be compatible with any work on that aspect from your side. We agree
a rule partition is a good point, but also that the way to do so can become
quite complex as you improve the cost analysis function. We can benefit too
from such advances.

Our proposal at this moment is just for "sequence of events". Afterwards, we
will think about extending it to other cases, involving operations. Of
course, in this later case, to make operational sense we won't be able to
trigger an event per event, but some kind of "micro window with a report"
will have to be applied.
 
Also, at a later iteration we could work on reducing the triggered events
with some intelligence and knowledge of the underlying structure. For
example, if the next rule depends on a dimension that is partitioned to the
same partition and the current rule, then a trigger would not be needed.

But the problem here is that this kind of tricks might be dependent on the
transport mechanism (fe. Kafka).


I would be very grateful if you would help me with at which Siddhi Core
classes I should implement the distribution logic. Where are the states
stored? Which class does the state handling?

Thank you again.

Regards,
Pablo



--
View this message in context: 
http://wso2-oxygen-tank.10903.n7.nabble.com/PROPOSAL-Distribute-internal-machine-states-tp144849p144902.html
Sent from the WSO2 Architecture mailing list archive at Nabble.com.
_______________________________________________
Architecture mailing list
[email protected]
https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture

Reply via email to