Hi On Thu, Mar 12, 2015 at 8:30 PM, Andreas Tille <[email protected]> wrote:
> Hi Akshita, > > I think your hint was of great help! Droping the GROUP BY increased the > speed by an order of magnitude. I also droped the other similarly > redundant GROUP BY (see my last commit). This has caused a drastical > performance kick. if you compare the old result: > > I had noticed the redundant GROUP BY in the two sub-queries too. I removed them, and ran the query and noticed that there was a vast improvement in the performance. I was just waiting for your confirmation about GROUP BY.. before concluding anything. Nevertheless, I am glad, I proved to be of some help. > I think this would be also acceptable to run on production UDD. > However, I wonder whether you could do some further review after this > issue should be clarified. Perhaps we get another factor 2 or 4 for > speeding up to make sure we will not uselessly stress test Debian > hardware. > > I am working on speeding up the query. Will update you on it soon. > In summary: Thanks a lot for your help to > a) Get bugs_udd working at all > b) Speed it up to some acceptable state. > That's really welcome since my time capacity is to hard occupied > by other things and I have obviously not given enough love to these > tools. > Thanks a lot for your guidance on this. I really find working on this project very learning and enriching. Another hint: In the dir misc/sql/ in the Git repository I have stored > some test queries. The script bugs.sh (or > bugs_including_metapackage_bugs.sh - I do not remember properly) was > used as template to test the queries quickly. If you want to use these > feel free to change these scripts - they have no other purpose than > testing. Make sure you check whether they are in sync with bugs_udd > first! > > I'll definitely look into this. Thanks again, Akshita Jha
