Odysseus

This is the official support forum for Odysseus

Odysseus Benchmarking

User avatar
Marco Grawunder
Posts: 255
Joined: Tue Jul 29, 2014 10:29 am
Location: Oldenburg, Germany
Contact:

Re: Odysseus Benchmarking

Postby Marco Grawunder » Thu Feb 16, 2017 12:35 pm

But the latency is always lend-minlstart and could not be changed to lend-maxlstart, correct?


It is now selectable in the view.

stefan
Posts: 85
Joined: Tue Jul 12, 2016 1:03 pm

Re: Odysseus Benchmarking

Postby stefan » Thu Feb 16, 2017 9:04 pm

Hi Marco,

Thanks for your effort! Ok, I did some tests and now:

According to the 1st issue (min/max latency):
Ok, true. But if I aggregate the tuples of a stream and join these aggregated tuples to the initial stream again, is the min latency be changed through the aggregation? Or is it still the min latency of all aggregated tuples, e. g. min(minlstart[i]) where i is loop counter?

Thanks for the option to use max latency also, I will test that.

According to the 2nd issue (RabbitMQ Sender):
I tested this and I got some errors. I assume this results because of the structure:
1. Read CSV
2. Do some preparation
3. Use Tupletokeyvalue operator
4. Send the keyvalue objects to RabbitMQ

I just get empty latency files (the csv is created but no values are inserted) but this errors:

Code: Select all

java.lang.ClassCastException: de.uniol.inf.is.odysseus.keyvalue.datatype.KeyValueObject cannot be cast to de.uniol.inf.is.odysseus.core.collection.Tuple
   at de.uniol.inf.is.odysseus.core.datahandler.TupleDataHandler.writeData(TupleDataHandler.java:232)
   at de.uniol.inf.is.odysseus.core.datahandler.AbstractStreamObjectDataHandler.writeData(AbstractStreamObjectDataHandler.java:149)
   at de.uniol.inf.is.odysseus.core.datahandler.AbstractStreamObjectDataHandler.writeCSVData(AbstractStreamObjectDataHandler.java:200)


I guess this occurs during reading the latency of the tuple - but its no tuple at this time.
If this results in a lot of work for you, just inform me. This query is also easy to measure manually. :)

According to the 3rd issue (standalone evaluation does not work):
There is no difference. I checked the features: Evaluation, Latency, Datarate + Systemload metadata, Systemload metadata are installed.
I got no errors but there is noting executed also.

greetings,
Stefan

User avatar
Marco Grawunder
Posts: 255
Joined: Tue Jul 29, 2014 10:29 am
Location: Oldenburg, Germany
Contact:

Re: Odysseus Benchmarking

Postby Marco Grawunder » Thu Feb 16, 2017 9:52 pm

Ok, true. But if I aggregate the tuples of a stream and join these aggregated tuples to the initial stream again, is the min latency be changed through the aggregation? Or is it still the min latency of all aggregated tuples, e. g. min(minlstart[i]) where i is loop counter?


The min latency of an aggregation is defined by the element with the min latency in this aggreation (typically the last one).

According to the 2nd issue (RabbitMQ Sender):


Yes ... I know what the reason is ... Currently, the approach does not work, when the sinks are KeyValueBased ... I can fix this next week. Not this week anymore...

The Evaluation feature works for me standalone ... Just to be sure: The queries alone are working?

stefan
Posts: 85
Joined: Tue Jul 12, 2016 1:03 pm

Re: Odysseus Benchmarking

Postby stefan » Fri Feb 17, 2017 12:39 pm

Hi Marco,

The min latency of an aggregation is defined by the element with the min latency in this aggreation (typically the last one).


Thanks, thats what I wanted to know/to get confirmed. :)

Yes ... I know what the reason is ... Currently, the approach does not work, when the sinks are KeyValueBased ... I can fix this next week. Not this week anymore...


Yep, I thought this would be the problem.
Next week is totally fine. It would be also fine if it will be fixed later. I just have to know it to decide how to go ahead.

The Evaluation feature works for me standalone ... Just to be sure: The queries alone are working?


Yes, the queries alone are working. I only have to define the parameters of the evaluation by hand, then the queries are working.
But if it is working for you, its ok. I can do my evaluation in the eclipse version also.

Have a nice weekend,
Stefan

stefan
Posts: 85
Joined: Tue Jul 12, 2016 1:03 pm

Re: Odysseus Benchmarking

Postby stefan » Fri Feb 17, 2017 5:42 pm

Hallo Marco,

during my tests I pulled the current sources from git also. I noticed that my already successfully executed evaluation jobs are not running anymore.

In the console output I got the following error (and only this output):

Code: Select all

java.lang.ClassCastException: org.eclipse.core.internal.resources.Project cannot be cast to org.eclipse.core.resources.IFile
   at de.uniol.inf.is.odysseus.rcp.evaluation.execution.EvaluationJob.fileToLines(EvaluationJob.java:288)
   at de.uniol.inf.is.odysseus.rcp.evaluation.execution.EvaluationJob.run(EvaluationJob.java:79)
   at org.eclipse.core.internal.jobs.Worker.run(Worker.java:55)



Maybe this is also the reason why it is not running in the standalone version. To make sure that this is really a problem I pulled the sources again but the problem still exists...

greetings,
Stefan

User avatar
Marco Grawunder
Posts: 255
Joined: Tue Jul 29, 2014 10:29 am
Location: Oldenburg, Germany
Contact:

Re: Odysseus Benchmarking

Postby Marco Grawunder » Sun Feb 19, 2017 4:32 pm

Yes, I made an extension (for setup und teardown query, as in JUnit) but forgot a test ... should be fixed now.

User avatar
Marco Grawunder
Posts: 255
Joined: Tue Jul 29, 2014 10:29 am
Location: Oldenburg, Germany
Contact:

Re: Odysseus Benchmarking

Postby Marco Grawunder » Mon Feb 20, 2017 1:48 pm

Ok. The Evaluation feature should now work with key value, too.

In my tests, everthing works fine...


phpbb 3.1 style demo

Return to “Other Topics”

Who is online

Users browsing this forum: No registered users and 1 guest