Ignite - CEP

[M Singh <ma...@yahoo.com>]

Hi:
I am trying to understand how CEP works in Ignite and from my understanding, we need to create a sliding window with time (or some other criteria) based eviction policy.  At the expiration of the policy the entry will be removed from the cache.
I wanted to find out if there is a way to still keep the item in the cache but still allow a sliding window without using the eviction policy.  The scenario I have is that I would like to query say the top 5 items in last 10 minutes but sill be able to query the the query the previous windows.
Thanks
Mans

Re: Ignite - CEP

[vkulichenko <va...@gmail.com>]

Hi,

Here are my answers.

0. Not sure about production usage, but the described approach makes sense
to me.
1. Ignite is designed to be scaleable, so you just need to make sure that
you have enough memory for your data. Refer to [1] for information about
capacity planning.
2-3. You can simply deploy a node singleton service [2] and Ignite will
guarantee that there is always an instance of the service on each node. If a
new node joins, it will be deployed there automatically as well.

[1] https://apacheignite.readme.io/docs/capacity-planning
[2] https://apacheignite.readme.io/docs/service-grid

-Val



--
View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Ignite-CEP-tp6497p6568.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.

Re: Ignite - CEP

[barrettbnr <ba...@bossanova.com>]

We are preparing to deploy Ignite as part of CEP system.

We have the following components

- A loader using the DataStreamer to continually push into a single 
- Clients registering continuous queries
- Instead of using an expiration policy we are deploying a  service to
selectively delete entries. So the stream becomes progressively more sparse
over time. We call this the GCService
- We don't have any strong consistency requirements, entries are write-once. 
- All entries have an affinity key.


Questions

0. Any obvious issues with the above? In particular things people have seen
in production?
1. Regarding above response what would be considered "big frequency" for CEP
to handle?
2. What is the best way to deploy the service in an HA fashion to run as a
scheduled job? e.g. our goal is to say every 30 mins execute this Service
(which deletes entry) and have a per-node service.
3. If a new node is added, are per-node services automatically deployed onto
that node? If not what is the best programmatic way to do that, we can't do
it in config because we stay away from Spring.

Thanks!



--
View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Ignite-CEP-tp6497p6566.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.

Re: Ignite - CEP

[Andrey Gura <ag...@gridgain.com>]

Mans,

You can configure indexes for you entities that will take in account entity
creation time. Then you can periodically query cache data using SQL query
with time bounds (last 10 minutes). But be aware that in case of big
frequency of adding entries into the cache memory could be exhausted.

On Sun, Jul 24, 2016 at 8:44 PM, M Singh <ma...@yahoo.com> wrote:

> Hi:
>
> I am trying to understand how CEP works in Ignite and from my
> understanding, we need to create a sliding window with time (or some other
> criteria) based eviction policy.  At the expiration of the policy the entry
> will be removed from the cache.
>
> I wanted to find out if there is a way to still keep the item in the cache
> but still allow a sliding window without using the eviction policy.  The
> scenario I have is that I would like to query say the top 5 items in last
> 10 minutes but sill be able to query the the query the previous windows.
>
> Thanks
>
> Mans
>



-- 
Andrey Gura
GridGain Systems, Inc.
www.gridgain.com