Archived copy. This page was recovered from the Internet Archive snapshot of /pages/viewpage.action?pageId=41386428 taken on 2021-09-24. The original wiki at wiki.opf-labs.org is being decommissioned.

Evaluation1 - SB Experiment Audio mp3 to wav Migration and QA on Hadoop Cluster

Created by Bolette Ammitzbøll Jurik on Apr 10, 2014 · last edited by Bolette Ammitzbøll Jurik · on Apr 10, 2014 (view history) · 13 versions

Evaluator(s)

Bolette Jurik (SB)

Evaluation points

Assessment of measurable points

Metric	Description	Metric baseline	Metric goal
NumberOfObjectsPerHour	Performance efficiency - Capacity / Time behaviour	18 (9th-13th November 2012)	1000
NumberOfFailedFiles	Reliability - Runtime stability	0	0
QAFalseDifferentPercent	Functional suitability - Correctness	0.412 % (5th-9th November 2012)	0.412 %

Small Experiments

All run on a file list of 58 files.

max split size	duration	launched maps	success	failure
1024	37m, 58.593s = 2278.593s	3,3,7	18	40
512	24m, 1.9s = 1441.9s	6,6,14	0	58
256	18m, 17.917 = 1097.917	12,12,28	0	58
128	17m, 3.176 = 1023.176	24,24,57	10	48
64	16m, 54.703s = 1014.703s	47,47,113	0	58
32	17m, 29.96s = 1049.96	93,93,225	4	54

The big question is why we get so many failures? The answer is of course that the list of pairs of files to compare is wrong! This list is created by Taverna beanshells, and we are missing a sort of the two output lists from the FFmpeg and mpg321 Hadoop jobs, before we combine the lists to a list of pairs as input to the waveform-compare Hadoop job. This should be a fairly quick fix...

The exact number of MR maps seem not to have a big influence on performance, as long as we have more than 12. That is as long as max split size is at most 256. We note that we get approximately twice as many launched maps for the waveform-compare Hadoop job, simply because the input list is approximately twice as big, as it is a list of pairs. We can of course adjust this to get approximately the same number of jobs, but it does not seem to be important for the performance.

The first line of tests were to decide on expected optimal max split size.

The next line of tests will use 128 as max split size and vary on the size of the input

Assessment of non-measurable points

ReliableAndStableAssessment Reliability - Runtime stability

For some evaluation points it makes most sense to a textual description/explanation

Please include a note about goals-objectives omitted, and why.

Technical details

Remember to include relevant information, links, versions about workflow, tools, APIs (e.g. Taverna, command line, Hadoop, links to MyExperiment, link to tools or SCAPE name, links to distinct versions of specific components/tools in the component registry)

WebDAV

We would like to store sufficient information about an experiment (hadoop program, configuration, etc.), so we are able to rerun it. For this purpose, ONB is providing a WebDAV - if you have questions and need more information, please contact Sven or Reinhard at ONB.
Taverna workflows will still be stored on myexperiment.org.

Link: http://fue.onb.ac.at/scape-tb-evaluation

Please use the following structure for storing experiment results

Evaluation notes

Could be such things as identified issues, workarounds, data preparation, if not already included above

QAFalseDifferentPercent