Re: yet another q - Mailing list pgsql-performance

On Wed, Aug 18, 2010 at 11:14 PM, Samuel Gendler <sgendler@ideasculptor.com> wrote:

Please forgive the barrage of questions.  I'm just learning how to tune things in postgres and I've still got a bit of learning curve to get over, apparently.  I have done a lot of reading, though, I swear.

I've got two identical queries except for a change of one condition which cuts the number of rows in half - which also has the effect of eliminating one partition from consideration (partitioned over time and I cut the time period in half).  The query plans are considerably different as a result. The net result is that the fast query is 5x faster than the slow query.  I'm curious if the alternate query plan is inherently faster or is it just a case of the algorithm scaling worse than linearly with the row count, which certainly wouldn't be surprising.  The big win, for me, is that the sort uses vastly less memory.  The slow plan requires work_mem to be 1500MB to even make it 5x worse.  With a more reasonable work_mem (400MB), it drops to something like 15x worse because it has to sort on disk.

fast plan: http://explain.depesz.com/s/iZ
slow plan: http://explain.depesz.com/s/Dv2

query:


EXPLAIN ANALYZE SELECT
           t_lookup.display_name as group,
           to_char(t_fact.time, 'DD/MM HH24:MI') as category,          
           substring(t_lookup.display_name from 1 for 20) as label,
           round(sum(t_fact.total_ms)/sum(t_fact.count)) as value
       FROM
           portal.providers t_lookup,
           day_scale_radar_performance_fact t_fact
       WHERE
           t_fact.probe_type_num < 3
           and t_lookup.provider_id = t_fact.provider_id
           and t_lookup.provider_owner_customer_id = t_fact.provider_owner_customer_id
           and t_fact.provider_owner_customer_id = 0
           and t_fact.time between timezone('UTC', '2010-08-18 15:00:00') - interval '30 day' and timezone('UTC', '2010-08-18 15:00:00')
       GROUP BY
           t_fact.provider_owner_customer_id, t_fact.provider_id,
           t_lookup.display_name,
           t_fact.time
       ORDER BY
           t_fact.time

table structure:

       Table "perf_reporting.abstract_radar_performance_fact"
          Column           |            Type             | Modifiers
----------------------------+-----------------------------+-----------
count                      | bigint                      | not null
total_ms                   | bigint                      | not null
time                       | timestamp without time zone | not null
market_num                 | integer                     | not null
country_num                | integer                     | not null
autosys_num                | integer                     | not null
provider_owner_zone_id     | integer                     | not null
provider_owner_customer_id | integer                     | not null
provider_id                | integer                     | not null
probe_type_num             | integer                     | not null

with individual indexes on the everything from time to the bottom on the child tables

and


                            Table "portal.providers"
          Column           |            Type             |       Modifiers        
----------------------------+-----------------------------+------------------------
btime                      | timestamp without time zone | not null default now()
mtime                      | timestamp without time zone | not null default now()
version                    | integer                     | not null default 1
provider_id                | integer                     | not null
provider_owner_zone_id     | integer                     | not null
provider_owner_customer_id | integer                     | not null
provider_category_id       | integer                     | not null
name                       | character varying(255)      | not null
display_name               | character varying(255)      | not null

with indexes on every column with name ending in '_id'

It gets more complicated:

When I dropped to a query over 15 days instead of 30 days, I saw a huge bump in performance (about 16 secs), the query plan for which is here:

http://explain.depesz.com/s/iaf

note: the query is identical to the one below, but with the interval changed to 15 days from 30 days, which also keeps the query within a single partition.  Note that the sort requires almost no memory and occurs after the aggregation.  I thought my problems were solved, since reducing the normal window over which queries are performed is something the app can tolerate.

However, if I keep the same 15 day window (so row count is approximately the same), but change the time window start date by 2 days (still keeping the entire query within the same partition), I get a completely different query plan.  There is effectively no difference between the two queries other than the start date of the time window in the where clause, but one executes in twice the time (35 secs or thereabouts).

http://explain.depesz.com/s/LA

Just for completeness' sake, I changed the query such that it is still 15 days, but this time crosses a partition boundary.  The plan is very similar to the previous one and executes in about the same time (35 secs or so)

http://explain.depesz.com/s/Aqw

Statistics are up to date and were performed with default_statistics_target = 100

Is there any way I can force the more efficient HashAggregate then sort plan instead of sort then GroupAggregate?