Closed rju closed 1 week ago
author André van Hoorn -- Fri, 18 May 2012 09:54:41 +0200
Jan Waller: As mentioned, such ticket exists already
Till: Would allow to have
1. OPAD as a monolithic filter which is composed by the OPAD filters for specific functionality, like time series extraction, time series forecasting, anomaly score computation, anomaly detection, ...
1. Re-use these included filters outside OPAD
author Tillmann Carlos Bielefeld -- Fri, 18 May 2012 10:11:49 +0200
This would also be beneficial to provide composite structures as examples. Interested people could just download Kieker, run it, and copy the xml part of this composite structure into their .kax file (could we rename it to kix, btw?).
Also, this would allow to have certain common setups and would simplify the design process. Just my 2c.
author André van Hoorn -- Fri, 18 May 2012 10:21:58 +0200
Replying to [tcb|comment:2]:
> This would also be beneficial to provide composite structures as examples. Interested people could just download Kieker, run it, and copy the xml part of this composite structure into their .kax file
Sure, an example would be needed. However, I don't see .kax files to be an artifact to be edited manually.
> (could we rename it to kix, btw?).
.kax was the result of a democratic process ( KIEKER-544 Done ). Seriously, I really don't see a problem with this name. OK, in Germany it may cause some laughter, but this is what names of well-known conferences like WICSA or ICAC do as well.
author Tillmann Carlos Bielefeld -- Fri, 18 May 2012 10:37:21 +0200
> cause some laughter
Alright, I am going to see the benefits in that! It was just my last try, don't bother
- I'm fine with it
> However, I don't see .kax files to be an artifact to be edited manually.
Mh, how could we provide composite structures then? Or do you think that's infeasible in general? Maybe we should just keep that in mind and stop the discussion here till it comes up at implementation time.
author André van Hoorn -- Fri, 18 May 2012 10:41:58 +0200
> > However, I don't see .kax files to be an artifact to be edited manually.
> Mh, how could we provide composite structures then? Or do you think that's infeasible in general? Maybe we should just keep that in mind and stop the discussion here till it comes up at implementation time.
Would be my suggestion, right.
What I had in mind was that the composite structure is still implemented in Java using the Plugin annotations etc., but an abstract super class provides means to register the filters and connections within this composite structure rather than in the top-level analysis controller. But let's see.
author nils-christian -- Mon, 22 Apr 2013 22:47:44 +0200
As discussed with jwa today: This is potential part of my Master' thesis.
author André van Hoorn -- Mon, 18 Nov 2013 15:48:41 +0100
Possible use cases
author André van Hoorn -- Mon, 24 Feb 2014 13:24:05 +0100
Nils addressed this in his master's thesis. At least the API is worth to look at.
See Nils' thesis and branch distributedAnalysis
author Thomas F. Düllmann -- Fri, 2 May 2014 16:52:54 +0200
The implementation from the development project is included in the opad-tslib-integration branch (changeset:3bf24604fd23fc96d20fb1a7b063be42e58212aa/kieker-git) and should be refactored, e.g.:
1. tools/configuration should be moved to analysis/configuration
1. same for the JUnit tests (composite)
author rju -- Sun, 18 Feb 2018 20:02:18 +0100
TeeTime supports composite filters.
author rju -- Tue, 12 Feb 2019 13:09:44 +0100
The new TeeTime framework supports composite stages which addresses this issue. Therefore, we now have composite filters. We must specify potential composite filters and then create specific tickets for them to be able to hand out assignments to students. I've added a Student tag, as I could not find one. Do we have another tag for that?
author André van Hoorn -- Thu, 25 Apr 2019 09:50:40 +0200
Supported by TeeTime.
JIRA Issue: KIEKER-1094 Support for composite filters in Kieker Original Reporter: Andre van Hoorn
Just came into my mind. Via dirty hacks (creating an AnalysisController inside a filter ...) this is already possible. However, framework support might be nice, e.g., via another abstract class AbstractCompositeAnalysisFilter extending AbstractAnalysisFilter.