Postgresql – Postgres – how to avoid unnecessary comparisons inside a WHERE with a JOIN

countjoin;performancepostgresqlpostgresql-9.6query-performance

Background

We're running this query:

SELECT COUNT(*) AS "__count"
FROM "xfiler_document"
INNER JOIN "xfiler_tx" ON ("xfiler_document"."tx_id" = "xfiler_tx"."id")
WHERE (
    "xfiler_document"."added_on" <= '2015-06-28T23:59:59.999999-04:00'::timestamptz
    AND "xfiler_tx"."company_id" = 1
    AND "xfiler_document"."added_on" >= '2015-06-01T00:00:00-04:00'::timestamptz
)

The result of EXPLAIN ANALYZE is @ https://explain.depesz.com/s/voMo

The table is large (+30 million records), so we've run into a performance problem with a report we're running (this can be seen at the URL above; +25 second query).

Question

It seems to me, based on the analyze results, that the performance issue here is due to the fact that Postgres is comparing the added_on values (step 4), and then re-checking them for some reason (step 3) for every row from the query's added_on time frame, rather than only checking the 18,515 rows that apparently match the company_id comparison in step 7.

Firstly, am I understanding / explaining the performance issue correctly?

Secondly, is there a way to solve this (other than denormalizing company_id)?

Database

Postgres 9.6.1

xfiler_document

       Column        |            Type             |                          Modifiers                           
---------------------+-----------------------------+--------------------------------------------------------------
 id                  | integer                     | not null default nextval('xfiler_document_id_seq'::regclass)
 tx_id               | integer                     | not null
 doc                 | character varying(255)      | 
 added_on            | timestamp with time zone    | not null

Indexes:
    "xfiler_document_pkey" PRIMARY KEY, btree (id)
    "xfiler_document_tx_id" btree (tx_id)
    "xfiler_document_added_on" btree (added_on)
Foreign-key constraints:
    "xfiler_document_tx_id_fkey" FOREIGN KEY (tx_id) REFERENCES xfiler_tx(id) DEFERRABLE INITIALLY DEFERRED

xfiler_tx

           Column           |           Type           |                       Modifiers                        
----------------------------+--------------------------+--------------------------------------------------------
 id                         | integer                  | not null default nextval('xfiler_tx_id_seq'::regclass)
 name                       | character varying(255)   | not null
 company_id                 | integer                  | not null

Indexes:
    "xfiler_tx_pkey" PRIMARY KEY, btree (id)
    "xfiler_tx_company_id" btree (company_id)
    "xfiler_tx_name" btree (name)
Foreign-key constraints:
    "xfiler_tx_company_id_fkey" FOREIGN KEY (company_id) REFERENCES company_company(id) DEFERRABLE INITIALLY DEFERRED
Referenced by:
    TABLE "xfiler_document" CONSTRAINT "xfiler_document_tx_id_fkey" FOREIGN KEY (tx_id) REFERENCES xfiler_tx(id) DEFERRABLE INITIALLY DEFERRED

Best Answer

The plan entry Recheck Cond: indicates what conditions will be used for the recheck, should a recheck be necessary. This will always be listed, even if no recheck is actually necessary at run time. (You can tell this because the Recheck Cond shows up even if you do a regular EXPLAIN, not an EXPLAIN ANALYZE, so this is part of the planning process, not the execution process.)

In your situation, all Heap Blocks were exact, meaning no recheck was necessary.

The real explanation for the slowness is probably that the Bitmap Heap Scan needs to read a lot of data from disk in order to find the rows it needs. You could check this by turning track_io_timing on and running EXPLAIN (ANALYZE, BUFFERS).

If that is indeed the issue, one solution would be to CLUSTER xfiler_document ON xfiler_document_added_on so that the data for a given date range are closer together on disk.

If your data is on a RAID, then increasing effective_io_concurrency could help.

Related Solutions

Postgresql – Postgres multiple joins slow query, how to store default child record

You write:

Each customer can have multiple sites, but only one should be displayed in this list.

Yet, your query retrieves all rows. That would be a point to optimize. But you also do not define which site is to be picked.

Either way, it does not matter much here. Your EXPLAIN shows only 5026 rows for the site scan (5018 for the customer scan). So hardly any customer actually has more than one site. Did you ANALYZE your tables before running EXPLAIN?

From the numbers I see in your EXPLAIN, indexes will give you nothing for this query. Sequential table scans will be the fastest possible way. Half a second is rather slow for 5000 rows, though. Maybe your database needs some general performance tuning?

Maybe the query itself is faster, but "half a second" includes network transfer? EXPLAIN ANALYZE would tell us more.

If this query is your bottleneck, I would suggest you implement a materialized view.

After you provided more information I find that my diagnosis pretty much holds.

The query itself needs 27 ms. Not much of a problem there. "Half a second" was the kind of misunderstanding I had suspected. The slow part is the network transfer (plus ssh encoding / decoding, possibly rendering). You should only retrieve 100 rows, that would solve most of it, even if it means to execute the whole query every time.

If you go the route with a materialized view like I proposed you could add a serial number without gaps to the table plus index on it - by adding a column row_number() OVER (<your sort citeria here>) AS mv_id.

Then you can query:

SELECT *
FROM   materialized_view
WHERE  mv_id >= 2700
AND    mv_id <  2800;

This will perform very fast. LIMIT / OFFSET cannot compete, that needs to compute the whole table before it can sort and pick 100 rows.

pgAdmin timing

When you execute a query from the query tool, the message pane shows something like:

Total query runtime: 62 ms.

And the status line shows the same time. I quote pgAdmin help about that:

The status line will show how long the last query took to complete. If a dataset was returned, not only the elapsed time for server execution is displayed, but also the time to retrieve the data from the server to the Data Output page.

If you want to see the time on the server you need to use SQL EXPLAIN ANALYZE or the built in Shift + F7keyboard shortcut or Query -> Explain analyze. Then, at the bottom of the explain output you get something like this:

Total runtime: 0.269 ms

Postgresql – Simple PostgreSQL lookup table is inexplicably slow

Rather than "VARCHAR(40000)" why not use "TEXT"?
HASH index use is discouraged (see docs at http://www.postgresql.org/docs/current/static/indexes-types.html).
Have you run "ANALYZE" on your tables before running the query?
Giant IN lists can be performance killers.

How do the following queries perform?

SELECT e.entityId
FROM entities e
INNER JOIN triples t ON (t.object = e.entityId)
LIMIT 10000;

SELECT e.entityId
FROM entities e
WHERE EXISTS (SELECT 1 FROM triples t WHERE t.object = e.entityId LIMIT 10000);