Postgresql – How to deal with fragmentation in PostgreSQL

clustered-indexperformancepostgresql

What are best practices for dealing with fragmentation in PostgreSQL?

Example:

I have a multi-tenant database whose workload includes on-the-fly aggregation for reports.

CREATE TABLE account (
    id serial PRIMARY KEY,
    name text NOT NULL
);

CREATE TABLE widget (
    id bigserial PRIMARY KEY,
    account_id int NOT NULL REFERENCES account (id),
    length numeric NOT NULL,
    weight numeric NOT NULL,
    color text NOT NULL,
    shape text NOT NULL
);

CREATE INDEX ON widget (account_id);

I want to run account-scoped queries on widget, e.g.

SELECT color, count(*), sum(weight) AS weight
FROM widget
WHERE account_id = 42094 AND 3 < length
GROUP BY 1

Suppose there are 5,000 records in account and 100,000,000 records in widget.

Since PostgreSQL lacks clustered indices, an individual account's 20,000 widget records get scattered all over the entire table. With enough records in the table, PostgeSQL will have to read 20,000 pages for a query that under other circumstances would be very fast.

This would seem to be a common sort of situation…how do users deal with this?

Best Answer

CLUSTER widget USING widget_account_id_idx;

That will rewrite the table in index order, and your query will become faster.

There are down sides:

during CLUSTER, the table will be inaccessible
the order is not maintained, so you have to run CLUSTER everynow and then

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

Mysql – Partitioning MySQL for “expired” transactions to improve performance

If a lot of your queries have conditions similar to what you describe, e.g. range condition on the date column:

WHERE dateColumn >= (CURRENT_DATE() - INTERVAL 1 MONTH)

WHERE dateColumn >= '2012-01-01' 
  AND dateColumn < '2013-01-01'

it will be useful to define the primary key of the table as (dateColumn, tableAI), where tableAI is an auto incremented integer column.

For InnoDB tables, the above choice means that this index will be the clustered index of the table and any range (on the date) queries on the table will use it and will only scan the relevant part of the table.

If the HasExpired condition is more complex, e.g. some rows should be marked expired after a month, some others after a year, some after a few days, I would consider splitting the table(s) into 2 parts. One having only expired data and one having all data.

So, when some data expires, you'll simply delete them from the "not_expired" table. The data will already be on the other. When you search for not expired data, you'll be searching only one table (the smallest one).
You also have the benefit of adding different indexes on the tables (for example only the above suggestion for the CI in the "all-data" table and many indexes on the not-expired table, depending on your queries.) Without the split, you would need to have an HasExpired column and attach it to all your indexes (as first column), to have equally good indexes.

If the (not-expired) data is updated many times, it might be beneficial to have the split as "not-expired" and "expired" data. So you won't need to do any updates on the "expired" table. But when some data expires, you'll simply move them (instead of deleting) from the "not_expired" table to the "expired" one.

Best Answer

Related Solutions

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

pgAdmin timing

Mysql – Partitioning MySQL for “expired” transactions to improve performance

Related Question