[
https://issues.apache.org/jira/browse/GEARPUMP-23?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15281081#comment-15281081
]
Manu Zhang commented on GEARPUMP-23:
------------------------------------
Thanks, Kam.
I read this paper some while ago. It has a great summary of algebraic
properties of aggregate operations. For example, count is associative,
invertible and commutative, and thus fits for out-of-order processing with
accumulating and distracting. Other operations may have more restrictions and
harder to handle in a window aggregation. It's worthwhile to explore how to
implement the approach of the paper on Gearpump for the long term.
The shortcut to go now is to stick to the Beam model, although my impression
(which could be wrong) is that it requires the aggregation to be associative
and commutative.
> Add DSL window (time series) support
> -------------------------------------
>
> Key: GEARPUMP-23
> URL: https://issues.apache.org/jira/browse/GEARPUMP-23
> Project: Apache Gearpump
> Issue Type: New Feature
> Components: streaming
> Affects Versions: 0.8.0
> Reporter: Kam Kasravi
> Assignee: Manu Zhang
> Fix For: 0.8.1
>
>
> Time series support is need for both GEARPUMP-21 and GEARPUMP-22. This was
> originally github issue
> [1917|https://github.com/gearpump/gearpump/issues/1917]. Other realtime
> streaming engines support time windows such as
> [Flink|https://flink.apache.org/news/2015/12/04/Introducing-windows.html],
> [Beam|https://www.oreilly.com/ideas/the-world-beyond-batch-streaming-101] and
> a future release of
> [Spark|https://issues.apache.org/jira/secure/attachment/12793410/StructuredStreamingProgrammingAbstractionSemanticsandAPIs-ApacheJIRA.pdf].
> A comparison matrix is found
> [here|https://docs.google.com/spreadsheets/d/1OM077lZBARrtUi6g0X0O0PHaIbFKCD6v0djRefQRE1I/edit#gid=995660187]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
