On Tue, Aug 27, 2013 at 10:18PM, Jay Vyas wrote:
> hmm ok. now. Thinking about teragen makes me think of benchmarking..
> In the longer term we could add benchmarking jobs to all the submodules not
> just mapreduce. For example there are hi bench and ycsb workloads which
> might be usable or pulled in as bigtop components ... Iff of course
> benchmarking is in the cards for bigtop?
It indeed is!
I think doing tera-gen/sort so it can be parameterized will provide a good
basis for future benchmarking (as a bit of reflection: I have did simplistic
yet efficient way of benchmarking HDFS and MR a couple years ago, but my
employer back then has never let it go into the open. Go figure...)
And I have a way of building YCSB against a particular version of Hadoop, so I
guess I will have it packaged as a benchmarking test pretty soon.
> On Aug 27, 2013, at 7:39 PM, Roman Shaposhnik <[EMAIL PROTECTED]> wrote:
> > On Tue, Aug 27, 2013 at 4:26 PM, Jay Vyas <[EMAIL PROTECTED]> wrote:
> >> Hi guys:
> >> I run TeraSort/TeraGen as additions to bigtop in some shell scripts.
> >> Any interest in these as an update to TestHadoopExamples in the MapReduce
> >> smokes?
> >> If so I could patch them in :)
> > Sure! Sounds like a useful addition.
> > Thanks,
> > Roman.