> On Dec 3, 2016, at 3:57 PM, Samuel Williams <space.ship.traveller@gmail.com> wrote:
>
> Thanks everyone for your feedback so far. I've done a bit more digging:
>
> MySQL in MBytes (about 350 million rows):
>
> index_user_event_on_what_category_id_created_at_latlng | 22806.00
> index_user_event_for_reporting | 18211.00
> index_user_event_on_created_at | 9519.00
> index_user_event_on_user_id | 6884.00
> index_user_event_on_poi_id | 4891.00
> index_user_event_on_deal_id | 3979.00
>
> Postgres (about 250 million rows):
>
> index_user_event_on_what_category_id_created_at_latlng | 25 GB
> index_user_event_for_reporting | 19 GB
> index_user_event_on_created_at | 7445 MB
> index_user_event_on_user_id | 7274 MB
> index_user_event_on_deal_id | 7132 MB
> index_user_event_on_poi_id | 7099 MB
>
> So, the index is a bit bigger, plus there is also the PKEY index which
> increases disk usage by another whole index. Keep in mind in the
> above, MySQL has about 40% more data.
>
> With some indexes, it looks like MySQL might not be adding all data to
> the index (e.g. ignoring NULL values). Does MySQL ignore null values
> in an index? Can we get the same behaviour in Postgres to minimise
> usage? What would be the recommendation here?
It's unlikely anyone will be able to usefully answer the questions you
should be asking without seeing the schema and index definitions,
and maybe some clues about how you're querying the data.
Cheers,
Steve
