[jira] [Commented] (QUARKS-192) event resiliency

["Daniel John Debrunner (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/QUARKS-192?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15312937#comment-15312937 ] 

Daniel John Debrunner commented on QUARKS-192:
----------------------------------------------

There is a mechanism to maintain tuples in memory for short outages.

https://github.com/apache/incubator-quarks/blob/master/api/topology/src/main/java/quarks/topology/plumbing/PlumbingStreams.java#L209

Basically put it in front of the IoT device and it will maintain the last N tuples and send then when the connection is restablished.

May not be exactly what you are looking for, but is useful for short outages or when only the most recent data is important. E.g. what good is it to know where you were two hours ago.

> event resiliency
> ----------------
>
>                 Key: QUARKS-192
>                 URL: https://issues.apache.org/jira/browse/QUARKS-192
>             Project: Quarks
>          Issue Type: Improvement
>            Reporter: Jerome Chailloux
>
> Assuming that I have a Raspberry Pi that ingest data and doing some aggregate.
> I have to send information to Watson IoT.
> What if if I lost the network ?
> Could we have a mechanism that send data at all time and if network is not available store all events in a file ?
> As soon as the network is available again play back the file and send new events.
> We have to be sure that all events are sent in the correct order from file to "real time".
> The computation engine (IBM Streams in this case) will then apply some policies to deal with or not with out of order events for the real time computation but at least able to store data received from sensors.
> Any idea about how to achieve this ?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)