Request for a branch on external hash groupby

[GoogleCodeExporter]

May I start a new branch for the external hash group operator? A test client 
will be added for performance test, and also possible more features (DISTINCT). 

Original issue reported on code.google.com by jarod...@gmail.com on 3 Feb 2011 at 4:20

[GoogleCodeExporter]

Jarod,

What is the folder structure you have in mind for the work you are going to do? 
I am guessing you are going to need to branch trunk so at the least you will 
have the hyracks code tree. I need clarification on where you intend to put the 
benchmarking client. I would like to see the benchmark client written in a way 
that it will go back to trunk and in addition to serving as a test case, will 
be an example for benchmarking operators. Once we close on that, let's create a 
branch.

Original comment by vinay...@gmail.com on 3 Feb 2011 at 5:06

[GoogleCodeExporter]

Hi Vinayak,

I am supposing to add one test client into the textclient package, and the 
DISTINCT operator into the dataflow.std.aggregator. 

For the bnechmarking package, based on your reply, I think you are looking for 
a new application (package) only for benchmarking external group operator (and 
also other implementations with other operators like external sortor, 
precluster)? If not, I think adding a client class in the textclient package is 
enough, since all aggregators have been included into the textapp.  

Best,

Original comment by jarod...@gmail.com on 3 Feb 2011 at 5:31

[GoogleCodeExporter]

Jarod,

I was expecting the benchmarking client to be more than just the code that 
invokes the different groupby operators. We would need a data generator to 
generate large amounts  of input data with different statistical properties to 
measure the efficiency of the operator.

Original comment by vinay...@gmail.com on 3 Feb 2011 at 5:41

[GoogleCodeExporter]

Branch created (r327): /branches/hyracks_spilling_groupby_perf

Original comment by vinay...@gmail.com on 6 Feb 2011 at 6:30

• Changed state: Done