HI Mohammad,

You're looking at the documentation for a different product - gemfirexd.
Check out the docs for geode here - http://geode.docs.pivotal.io/docs/.
GemfireXD is a SQL database built on top of gemfire (now geode).

Regarding your question as it applies to geode - I think using geode
functions is probably a good idea. With the cache listeners - you need to
be a bit careful about cache listeners if one listener is going to trigger
events that trigger's another listener and so on. You can get yourself in a
distributed deadlock if you create this long chain of listeners. So you
might be better of applying your original update in a function that
computes the side effects and fires other functions.

-Dan

On Wed, May 4, 2016 at 5:56 AM, Mohammad Adnan Raza <[email protected]>
wrote:

> One of my current assignment needs to implement real time graph processing
> and I am considering Geode to achieve this. We need to have a chain of data
> updates in a transaction on an record update event. Would it work if we
> enable a synchronous cache listener (as described here
> <http://gemfirexd.docs.pivotal.io/1.3.1/userguide/caching_database/writers-listeners-about.html>)
> on a record update event of a given table and call a data aware procedure
> (as described here
> <http://gemfirexd.docs.pivotal.io/1.3.1/userguide/developers_guide/topics/server-side/data-aware-procedures.html>).
> This procedure would execute a record put on a different table and given
> that it can execute on the node where the data resides, we would have
> better performance otherwise. Does this theoretically sound right?
>
> A detailed example is as following.
>
> Assume that I have node1, node2 and node3 all are integers. First event
> will come from user to update node1 with 10. From Graph I would know that
> next node is node2 and that should be updated by node1*2 i.e. 20. I'll
> update node2. Now node3 should be updated by node2*3 i.e. 60 I'll update
> node3.
>
> I would appreciate your thoughts on this.
> --
>
> *With Best Regards,*
>
> *       Mohammad Adnan*
>
>
>
>

Reply via email to