Postgresql - one ini4 vs two int2 for indexing - postgresql

I have got a table with millions of rows in postgresql. One row can be represent by eight int4 or sixteen int2 columns.
I want to have one multicolumn (btree) index on this table: create index on mytable(c1,c2,c3,c4,....c8);
I wonder, what is better solution (for performance purpose): one multicolumn index with eight (int4 type) columns or one multicolumn index with sixteen (int2 type) columns.
In other words:
create index on mytable (c_int4_1, c_int4_2, ... c_int4_8);
vs.
create index on mytable (c_int2_1,c_int2_2...c_int2_16);

Whichever most naturally matches the use of the data. Any gains from the more efficient on the btree will be lost again when forcing it into another format.

Related

Performance Difference between index on two columns vs conditional index

I have a decently large postgres table with a few billion rows.
However the table could be partitioned by one column (type)
Should we prefer:
An index with two columns
create nonclustered index ix_index1 on table1(type, string_urn_id)
or a conditional index
create nonclustered index ix_index1_alternative on table1(string_urn_id) WHERE type = 'type1'
create nonclustered index ix_index1_alternative2 on table1(string_urn_id) WHERE type = 'type2'
create nonclustered index ix_index1_alternative3 on table1(string_urn_id) WHERE type = 'type3'
....
There is no statement create nonclustered index in PostgreSQL.
What is better depends on the definition of "better". From a maintenance perspective, the single index is better, because you won't have to create a new index whenever you add a new type.
From a performance perspective, only a benchmark with realistic data can tell. Planning time will increase with many indexes, but query performance may be a tad better.
If you partition the table, query performance will decrease, but you can do with a single partitioned index on string_urn_id.

Indexing on Timescaledb

I am testing some queries on Postgresql extension Timescaledb.
The table is called timestampdb and i run some queries on that seems like this
select id13 from timestampdb where timestamp1 >='2010-01-01 00:05:00' and timestamp1<='2011-01-01 00:05:00',
select avg(id13)::numeric(10,2) from timestasmpdb where timestamp1>='2015-01-01 00:05:00' and timestamp1<='2015-01-01 10:30:00'
When i create a hypertable i do this.
create hyper_table('timestampdb','timestamp1')
The thing is that now i want to create an index on id13.
should i try something like this?:
create hyper_table('timestampdb','timestamp1') ,import data of the table and then create index on timestampdb(id13)
or something like this:
create table timestampdb,then create hypertable('timestampdb',timestamp1') ,import the data and then CREATE INDEX ON timestampdb (timestamp1,id13)
What is the correct way to do this?
You can create an index without time dimension column, since you don't require it to be unique. Including time dimension column into an index is needed if an index contains UNIQUE or is PRIMARY KEY, since TimescaleDB partitions a hypertable into chunks on the time dimension column, which is timestamp1 in the question. If partitioning key will include space dimension columns in addition to time, they will need to be included too.
So in your case the following should be sufficient after the migration to hypertable:
create index on timestampdb(id13);
The question contains two queries and none of them need index on id13. It will be valuable to create the index on id13 if you expect different queries than in the question, which will contain condition or join on id13 column.

Ideal postgres index for non unique varchar column

I need to create a varchar category column in a table and search for rows that are belonging to a particular category.
ie. ALTER TABLE items ADD COLUMN category VARCHAR(30)
The number of categories is very small (repeated across the table)
and the intention is to only use = in the where clause.
ie. select * from items where category = 'food'
What kind of index would be ideal in postgres?
Especially if the table is never expected to be too big (less than 5,000 rows always)
This is a textbook usecase for a Hash Index - you have a very small number of distinct values and only use the equality operator to query them. Using a hash index will enable you to index a relatively small hash of the value, which will allow for faster querying.

Index to query sorted values in keyed time range

Suppose I have key/value/timerange tuples, e.g.:
CREATE TABLE historical_values(
key TEXT,
value NUMERIC,
from_time TIMESTAMPTZ,
to_time TIMESTAMPTZ
)
and would like to be able to efficiently query values (sorted descending) for a specific key and time, e.g.:
SELECT value
FROM historical_values
WHERE
key = [KEY]
AND from_time <= [TIME]
AND to_time >= [TIME]
ORDER BY value DESC
What kind of index/types should I use to get the best lookup performance? I suspect my solution will involve a tstzrange and a gist index, but I'm
not sure how to make that play well with the key matching and value ordering requirements.
Edit: Here's some more information about usage.
Ideally uses features available in Postgres v9.6.
Relation will contain approx. 1k keys and 5m values per key. Values are large integers (up to 32 bytes), mostly unique. Time ranges between few hours to a couple years. Time horizon is 5 years. No NULL values allowed, but some time ranges are open-ended (could either use NULL or a time far into the future for to_time).
The primary key is the key and time range (as there is only one historical value for a time range, per key).
Common operations are a) updating to_time to "close" a historical value, and b) inserting a new value with from_time = NOW.
All values may be queried. Partitioning is an option.
DB design
For a big table like that ("1k keys and 5m values per key") I would suggest to optimize storage like:
CREATE TABLE hist_keys (
key_id serial PRIMARY KEY
, key text NOT NULL UNIQUE
);
CREATE TABLE hist_values (
hist_value_id bigserial PRIMARY KEY -- optional, see below!
, key_id int NOT NULL REFERENCES hist_keys
, value numeric
, from_time timestamptz NOT NULL
, to_time timestamptz NOT NULL
, CONSTRAINT range_valid CHECK (from_time <= to_time) -- or < ?
);
Also helps index performance.
And consider partitioning. List-partitioning on key_id. Maybe even add sub-partitioning on (range partitioning this time) on from_time. Read the manual here.
With one partition per key_id, (and constraint exclusion enabled!) Postgres would only look at the small partition (and index) for the given key, instead of the whole big table. Major win.
But I would strongly suggest to upgrade to at least Postgres 10 first, which added "declarative partitioning". Makes managing partition a lot easier.
Better yet, skip forward to Postgres 11 (currently beta), which adds major improvements for partitioning (incl. performance improvements). Most notably, for your goal to get the best lookup performance, quoting the chapter on partitioning in release notes for Postgres 11 (currently beta):
Allow faster partition elimination during query processing (Amit Langote, David Rowley, Dilip Kumar)
This speeds access to partitioned tables with many partitions.
Allow partition elimination during query execution (David Rowley, Beena Emerson)
Previously partition elimination could only happen at planning time,
meaning many joins and prepared queries could not use partition elimination.
Index
From the perspective of the value column, the small subset of selected rows is arbitrary for every new query. I don't expect you'll find a useful way to support ORDER BY value DESC with an index. I'd concentrate on the other columns. Maybe add value as last column to each index if you can get index-only scans out of it (possible for btree and GiST).
Without partitioning:
CREATE UNIQUE INDEX hist_btree_idx ON hist_values (key_id, from_time, to_time DESC);
UNIQUE is optional, but see below.
Note the importance of opposing sort orders for from_time and to_time. See (closely related!):
Optimizing queries on a range of timestamps (two columns)
This is almost the same index as the one implementing your PK on (key_id, from_time, to_time). Unfortunately, we cannot use it as PK index. Quoting the manual:
Also, it must be a b-tree index with default sort ordering.
So I added a bigserial as surrogate primary key in my suggested table design above and NOT NULL constraints plus the UNIQUE index to enforce your uniqueness rule.
In Postgres 10 or later consider an IDENTITY column instead:
Auto increment table column
You might even do with PK constraint in this exceptional case to avoid duplicating the index and keep the table at minimum size. Depends on the complete situation. You may need it for FK constraints or similar. See:
How does PostgreSQL enforce the UNIQUE constraint / what type of index does it use?
A GiST index like you already suspected may be even faster. I suggest to keep your original timestamptz columns in the table (16 bytes instead of 32 bytes for a tstzrange) and add key_id after installing the additional module btree_gist:
CREATE INDEX hist_gist_idx ON hist_values
USING GiST (key_id, tstzrange(from_time, to_time, '[]'));
The expression tstzrange(from_time, to_time, '[]') constructs a range including upper and lower bound. Read the manual here.
Your query needs to match the index:
SELECT value
FROM hist_values
WHERE key = [KEY]
AND tstzrange(from_time, to_time, '[]') #> tstzrange([TIME_FROM], [TIME_TO], '[]')
ORDER BY value DESC;
It's equivalent to your original.
#> being the range contains operator.
With list-partitioning on key_id
With a separate table for each key_id, we can omit key_id from the index, improving size and performance - especially for the GiST index - for which we then also don't need the additional module btree_gist. Results in ~ 1000 partitions and the corresponding indexes:
CREATE INDEX hist999_gist_idx ON hist_values USING GiST (tstzrange(from_time, to_time, '[]'));
Related:
Store the day of the week and time?

Should this PostgreSQL query use the indexes?

I have two tables:
CREATE TABLE soils (
sample_id TEXT PRIMARY KEY,
project_id TEXT,
technician_id TEXT
);
CREATE INDEX soils_idx
ON soils
USING btree
(sample_id COLLATE pg_catalog."default");
CREATE TABLE assays (
sample_id TEXT PRIMARY KEY,
mo_ppm NUMERIC
);
CREATE INDEX assays_idx
ON assays
USING btree
(sample_id COLLATE pg_catalog."default");
Each table contains about a half million records, and, in reality, about 20 additional columns each, of type TEXT (omitted in the DDL posted above to save time here).
When I perform the query:
EXPLAIN SELECT
s.sample_id, s.project_id, s.technician_id, a.mo_ppm
FROM
soils AS s INNER JOIN assays AS a ON s.sample_id = a.sample_id
I get 2 SEQ SCANs, rather than a lookup to the index. Is that expected behaviour?
Since you have no WHERE conditions, you effectively read the whole table. It's cheaper to run sequential scans and not involve any indexes at all.
Try:
EXPLAIN
SELECT s.sample_id, s.project_id, s.technician_id, a.mo_ppm
FROM soils s
JOIN assays a USING (sample_id)
WHERE <some condition that returns few rows>;
... and an index matching the WHERE condition should be used.
You don't need to define an index on a PRIMARY KEY column. A PK constraint is implemented with a unique index automatically. Your additional index is redundant and of no use.
An index on a foreign key column would be a good idea, but there isn't one in your example, which looks odd. Like the two tables could be combined into one. Probably just over-simplification for the test case.
Finally, for big tables, I would consider using a simple integer primary key instead of text, possibly a serial column. That's typically faster.
Yes, that's expected behaviour. On the other hand it depends on your random_page_cost, seq_page_cost and effective_cache_size settings. Your query doesn't have WHERE clause hence it might be faster to read everything sequentially. You can try to penalise sequential scan:
set enable_seqscan = off;
explain analyse <your query>;
and then compare plan/cost/IO wait (it is not possible to disable seq-scan but it gets very high cost -- ~1e7 (or 1e8)).
If you have SSD and WHERE clause in your query then you can lower random_page_cost to 1.5..2.5 and encourage PG to use index.