Re: [HACKERS] TPC-H Q20 from 1 hour to 19 hours!

Hi,

On 6/11/17 7:54 PM, Peter Geoghegan wrote:
> On Sun, Jun 11, 2017 at 10:36 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Do you mean teaching the optimizer to do something like this?:
>>
>> Uh, no.  I don't think we want to add any run-time checks.  The point in
>> this example is that we'd get a better rowcount estimate if we noticed
>> that the FK constraint could be considered while estimating the size of
>> the partsupp-to-aggregated-subquery join.
> 
> Sorry for not considering the context of the thread more carefully.
> Robert said something about selectivity estimation and TPC-H to me,
> which I decide to research; I then rediscovered this thread.
> 
> Clearly Q20 is designed to reward systems that do better with moving
> predicates into subqueries, as opposed to systems with better
> selectivity estimation.
> 

I do strongly recommend reading this paper analyzing choke points of 
individual TPC-H queries:
    http://oai.cwi.nl/oai/asset/21424/21424B.pdf

It's slightly orthogonal to the issue at hand (poor estimate in Q20 
causing choice of inefficient plan), it's a great paper to read. I 
thought I've already posted a link to the this paper sometime in the 
past, but I don't see it in the archives.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services