Hi Steve
Yes, benchmarking would be great. I kept thinking what
the best way to do a reasonable benchmark would be
(if we try it on a small data set caching will distort
the picture). My suggestion would be to use the
DC_W13_Stripe82_40deg, which is ~700 GB in size,
I'm going to create a copy of it on lsst10 and sort
relevant tables by objectId.
P.S. Dave has successfully finished running what he
needed over the weekend, but it'd still be a very
useful test.
thanks,
Jacek
On 02/21/2013 04:43 PM, Steve Bickerton wrote:
>
>> Helping Dave Monet
>> <snip>
>> - pipeQA probably operates ccd-at-a-time, sorting
>> the way we planned (by objectId) would help Dave
>> but would make pipeQA worse
>> --> need to check with pipeQA about access patterns
>
> PipeQA does indeed query on a ccd-at-a-time basis. I'm not sure how
> your proposed changes would affect pipeQa performance - or if they
> would, how severe a hit it would be. If you're interested in doing a
> benchmark test, let me know.
>
> s
>
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
|