When should I use a composite index in a database?
What are the performance ramification by using a
composite index)?
Why should I use use a composite index?
For example, I have a homes table:
CREATE TABLE IF NOT EXISTS `homes` (
`home_id` int(10) unsigned NOT NULL auto_increment,
`sqft` smallint(5) unsigned NOT NULL,
`year_built` smallint(5) unsigned NOT NULL,
`geolat` decimal(10,6) default NULL,
`geolng` decimal(10,6) default NULL,
PRIMARY KEY (`home_id`),
KEY `geolat` (`geolat`),
KEY `geolng` (`geolng`),
) ENGINE=InnoDB ;
Does it make sense for me to use a composite index for both geolat and geolng, such that:
I replace:
KEY `geolat` (`geolat`),
KEY `geolng` (`geolng`),
with:
KEY `geolat_geolng` (`geolat`, `geolng`)
If so:
Why?
What is the performance ramification by using a composite index)?
UPDATE:
Since many people have stated it entirely dependent upon the queries I perform, below is the most common query performed:
SELECT * FROM homes
WHERE geolat BETWEEN ??? AND ???
AND geolng BETWEEN ??? AND ???
UPDATE 2:
With the following database schema:
CREATE TABLE IF NOT EXISTS `homes` (
`home_id` int(10) unsigned NOT NULL auto_increment,
`primary_photo_group_id` int(10) unsigned NOT NULL default '0',
`customer_id` bigint(20) unsigned NOT NULL,
`account_type_id` int(11) NOT NULL,
`address` varchar(128) collate utf8_unicode_ci NOT NULL,
`city` varchar(64) collate utf8_unicode_ci NOT NULL,
`state` varchar(2) collate utf8_unicode_ci NOT NULL,
`zip` mediumint(8) unsigned NOT NULL,
`price` mediumint(8) unsigned NOT NULL,
`sqft` smallint(5) unsigned NOT NULL,
`year_built` smallint(5) unsigned NOT NULL,
`num_of_beds` tinyint(3) unsigned NOT NULL,
`num_of_baths` decimal(3,1) unsigned NOT NULL,
`num_of_floors` tinyint(3) unsigned NOT NULL,
`description` text collate utf8_unicode_ci,
`geolat` decimal(10,6) default NULL,
`geolng` decimal(10,6) default NULL,
`display_status` tinyint(1) NOT NULL,
`date_listed` timestamp NOT NULL default CURRENT_TIMESTAMP,
`contact_email` varchar(100) collate utf8_unicode_ci NOT NULL,
`contact_phone_number` varchar(15) collate utf8_unicode_ci NOT NULL,
PRIMARY KEY (`home_id`),
KEY `customer_id` (`customer_id`),
KEY `city` (`city`),
KEY `num_of_beds` (`num_of_beds`),
KEY `num_of_baths` (`num_of_baths`),
KEY `geolat` (`geolat`),
KEY `geolng` (`geolng`),
KEY `account_type_id` (`account_type_id`),
KEY `display_status` (`display_status`),
KEY `sqft` (`sqft`),
KEY `price` (`price`),
KEY `primary_photo_group_id` (`primary_photo_group_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci AUTO_INCREMENT=8 ;
Using the following SQL:
EXPLAIN SELECT homes.home_id,
address,
city,
state,
zip,
price,
sqft,
year_built,
account_type_id,
num_of_beds,
num_of_baths,
geolat,
geolng,
photo_id,
photo_url_dir
FROM homes
LEFT OUTER JOIN home_photos ON homes.home_id = home_photos.home_id
AND homes.primary_photo_group_id = home_photos.home_photo_group_id
AND home_photos.home_photo_type_id = 2
WHERE homes.display_status = true
AND homes.geolat BETWEEN -100 AND 100
AND homes.geolng BETWEEN -100 AND 100
EXPLAIN returns:
id select_type table type possible_keys key key_len ref rows Extra
----------------------------------------------------------------------------------------------------------
1 SIMPLE homes ref geolat,geolng,display_status display_status 1 const 2 Using where
1 SIMPLE home_photos ref home_id,home_photo_type_id,home_photo_group_id home_photo_group_id 4 homes.primary_photo_group_id 4
I don't quite understand how to read the EXPLAIN command. Does this look good or bad. Right now, I am NOT using a composite index for geolat and geolng. Should I be?
You should use a composite index when you are using queries that benefit from it. A composite index that looks like this:
index( column_A, column_B, column_C )
will benefit a query that uses those fields for joining, filtering, and sometimes selecting. It will also benefit queries that use left-most subsets of columns in that composite. So the above index will also satisfy queries that need
index( column_A, column_B, column_C )
index( column_A, column_B )
index( column_A )
But it will not (at least not directly, maybe it can help partially if there are no better indices) help for queries that need
index( column_A, column_C )
Notice how column_B is missing.
In your original example, a composite index for two dimensions will mostly benefit queries that query on both dimensions or the leftmost dimension by itself, but not the rightmost dimension by itself. If you're always querying two dimensions, a composite index is the way to go, doesn't really matter which is first (most probably).
Imagine you have the following three queries:
Query I:
SELECT * FROM homes WHERE `geolat`=42.9 AND `geolng`=36.4
Query II:
SELECT * FROM homes WHERE `geolat`=42.9
Query III:
SELECT * FROM homes WHERE `geolng`=36.4
If you have seperate index per column, all three queries use indexes. In MySQL, if you have composite index (geolat, geolng), only query I and query II (which is using the first part of the composit index) uses indexes. In this case, query III requires full table search.
On Multiple-Column Indexes section of manual, it is clearly explained how multiple column indexes work, so I don't want to retype manual.
From the MySQL Reference Manual page:
A multiple-column index can be
considered a sorted array containing
values that are created by
concatenating the values of the
indexed columns.
If you use seperated index for geolat and geolng columns, you have two different index in your table which you can search independent.
INDEX geolat
-----------
VALUE RRN
36.4 1
36.4 8
36.6 2
37.8 3
37.8 12
41.4 4
INDEX geolng
-----------
VALUE RRN
26.1 1
26.1 8
29.6 2
29.6 3
30.1 12
34.7 4
If you use composite index you have only one index for both columns:
INDEX (geolat, geolng)
-----------
VALUE RRN
36.4,26.1 1
36.4,26.1 8
36.6,29.6 2
37.8,29.6 3
37.8,30.1 12
41.4,34.7 4
RRN is relative record number (to simplify, you can say ID). The first two index generated seperate and the third index is composite. As you can see you can search based on geolng on composite one since it is indexed by geolat, however it's possible to search by geolat or "geolat AND geolng" (since geolng is second level index).
Also, have a look at How MySQL Uses Indexes manual section.
There could be a misconception about what composite index does. Many people think that composite index can be used to optimise a search query as long as the where clause covers the indexed columns, in your case geolat and geolng. Let's delve deeper:
I believe your data on the coordinates of homes would be random decimals as such:
home_id geolat geolng
1 20.1243 50.4521
2 22.6456 51.1564
3 13.5464 45.4562
4 55.5642 166.5756
5 24.2624 27.4564
6 62.1564 24.2542
...
Since geolat and geolng values hardly repeat itself. A composite index on geolat and geolng would look something like this:
index_id geolat geolng
1 20.1243 50.4521
2 20.1244 61.1564
3 20.1251 55.4562
4 20.1293 66.5756
5 20.1302 57.4564
6 20.1311 54.2542
...
Therefore the second column of the composite index is basically useless! The speed of your query with a composite index is probably going to be similar to an index on just the geolat column.
As mentioned by Will, MySQL provides spatial extension support. A spatial point is stored in a single column instead of two separate lat lng columns. Spatial index can be applied to such a column. However, the efficiency could be overrated based on my personal experience. It could be that spatial index does not resolve the two dimensional problem but merely speed up the search using R-Trees with quadratic splitting.
The trade-off is that a spatial point consumes much more memory as it used eight-byte double-precision numbers for storing coordinates. Correct me if I am wrong.
Composite indexes are useful for
0 or more "=" clauses, plus
at most one range clause.
A composite index cannot handle two ranges. I discuss this further in my index cookbook.
Find nearest -- If the question is really about optimizing
WHERE geolat BETWEEN ??? AND ???
AND geolng BETWEEN ??? AND ???
then no index can really handle both dimensions.
Instead, one must 'think out of the box'. If one dimension is implemented via partitioning and the other is implemented by carefully picking the PRIMARY KEY, one can get significantly better efficiency for very large tables of lat/lng lookup. My latlng blog goes into the details of how to implement "find nearest" on the globe. It includes code.
The PARTITIONs are stripes of latitude ranges. The PRIMARY KEY deliberately starts with longitude so that the useful rows are likely to be in the same block. A Stored Routine orchestrates the messy code for doing order by... limit... and for growing the 'square' around the target until you have enough coffee shops (or whatever). It also takes care of the great-circle calculations and handling the dateline and poles.
More
I have written another blog; it compares 5 ways of doing lat/lng searches: http://mysql.rjweb.org/doc.php/latlng#representation_choices (It references the link given above as one of the 5.) One of the other ways is this, and it points out that they are optimal for the particular case:
INDEX(geolat, geolng),
INDEX(geolng, geolat)
That is, having both columns in two indexes, and not having single-column indexes on geolat and geolng is important.
Composite indexes are very powerful as they:
Enforce structure integrity
Enable sorting on a FILTERED id
ENFORCE STRUCTURE INTEGRITY
Composite indexes are not just another type of index; they can provide NECESSARY structure to a table by enforcing integrity as the Primary Key.
Mysql's Innodb supports clustering and the following example illustrates why a composite index may be necessary.
To create a friends' tables (i.e. for a social network) we need 2 columns: user_id, friend_id.
Table Strcture
user_id (medium_int)
friend_id (medium_int)
Primary Key -> (user_id, friend_id)
By virtue, a Primary Key (PK) is unique and by creating a composite PK, Innodb will automatically check that no duplicates on user_id, friend_id exists when a new record is added. This is the expected behavior as no user should have more than 1 record (relationship link) with friend_id = 2 for instance.
Without a composite PK, we can create this schema using a surrogate key:
user_friend_id
user_id
friend_id
Primary Key -> (user_friend_id)
Now, whenever a new record is added we will have to check that a prior record with the combination user_id, friend_id does not already exist.
As such, a composite index can enforce structure integrity.
ENABLE SORTING ON A FILTERED ID
It is very common to sort a set of records by the post's time (timestamp or datetime). Usually, this means posting on a given id. Here is an example
Table User_Wall_Posts (think if Facebook's wall posts)
user_id (medium_int)
timestamp (timestamp)
author_id (medium_int)
comment_post (text)
Primary Key -> (user_id, timestamp, author_id)
We want to query and find all posts for user_id = 10 and sort the comment posts by timestamp (date).
SQL QUERY
SELECT * FROM User_Wall_Posts WHERE user_id = 10 ORDER BY timestamp DES
The composite PK enables Mysql to filter and sort the results using the index; Mysql will not have to use a temporary file or filesort to fetch the results. Without a composite key, this would not be possible and would cause a very inefficient query.
As such, composite keys are very powerful and suit more than the simple problem of "I want to search for column_a, column_b so I will use composite keys. For my current database schema, I have just as many composite keys as single keys. Don't overlook a composite key's use!
To do spacial searches, you need an R-Tree algorithm, which allows searching geographical areas very quickly. Exactly what you need for this job.
Some databases have spacial indexes built in. A quick Google search shows MySQL 5 has them (which looking at your SQL I'm guessing you're using MySQL).
Composite index can be useful when you want to optimise group by clause (check this article http://dev.mysql.com/doc/refman/5.0/en/group-by-optimization.html).
Please pay attention:
The most important preconditions for using indexes for GROUP BY are
that all GROUP BY columns reference attributes from the same index,
and that the index stores its keys in order (for example, this is a
BTREE index and not a HASH index)
There is no Black and White, one size fits all answer.
You should use a composite (or multi-column) index, when your query work load would benefit from one.
You need to profile your query work load in order to determine this.
A composite index comes into play when queries can be satisfied entirely from that index: meaning all the columns required by the query are in (covered) by an index.
UPDATE (in response to edit to posted question): If you are selecting * from the table the composite index may be used, it may not. You will need to run EXPLAIN PLAN to be sure.
I'm with #Mitch, depends entirely your queries. Fortunately you can create and drop indexes at any time, and you can prepend the EXPLAIN keyword to your queries to see if the query analyzer uses the indexes.
If you'll be looking up an exact lat/long pair this index would likely make sense. But you're probably going to be looking for homes within a certain distance of a particular place, so your queries will look something like this (see source):
select *, sqrt( pow(h2.geolat - h1.geolat, 2)
+ pow(h2.geolng - h1.geolng, 2) ) as distance
from homes h1, homes h2
where h1.home_id = 12345 and h2.home_id != h1.home_id
order by distance
and the index very likely won't be helpful at all. For geospatial queries, you need something like this.
Update: with this query:
SELECT * FROM homes
WHERE geolat BETWEEN ??? AND ???
AND geolng BETWEEN ??? AND ???
The query analyzer could use an index on geolat alone, or an index on geolng alone, or possibly both indexes. I don't think it would use a composite index. But it's easy to try out each of these permutations on a real data set and then (a) see what EXPLAIN tells you and (b) measure the time the query really takes.
Related
We have a large MySQL table (device_data) with the following columns:
ID (int)
dt (timestamp)
serial_number (char(20))
data1 (double)
data2 (double)
... // other columns
The table receives around 10M rows every day.
We have done a sharding by separating the table based on the date of the timestamp (device_data_YYYYMMDD). However, we feel this is not effective because most of our queries (shown below) always check on the "serial_number" and will perform across many dates.
SELECT * FROM device_data WHERE serial_number = 'XXX' AND dt >= '2018-01-01' AND dt <= '2018-01-07';
Therefore, we think that creating the sharding based on the serial number will be more effective. Basically, we will have:
device_data_<serial_number>
device_data_0012393746
device_data_7891238456
Hence, when we want to find data for a particular device, we can easily reference as:
SELECT * FROM device_data_<serial_number> WHERE dt >= '2018-01-01' AND dt <= '2018-01-07';
This approach seems to be effective because:
The application at all time will access the data based on the device first.
We have checked that there is no query that access the data without specifying the device serial number first.
The table for each device will be relatively small (9000 rows per day)
A few challenges that we think we will face is:
We have alot of devices. This means that the table device_data_ will be alot too. I have checked that MySQL does not provide limitation in the number of tables in the database. Will this impact on performance vs keeping them in one table?
How will this impact on later on when we would like to scale MySQL (e.g. using master / slave, etc)?
Are there other alternative / solution in resolving this?
Update. Below is the show create table result from our existing table:
CREATE TABLE `test_udp_new` (
`id` int(20) unsigned NOT NULL AUTO_INCREMENT,
`dt` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
`device_sn` varchar(20) NOT NULL,
`gps_date` datetime NOT NULL,
`lat` decimal(10,5) DEFAULT NULL,
`lng` decimal(10,5) DEFAULT NULL,
PRIMARY KEY (`id`),
KEY `device_sn_2` (`dt`,`device_sn`),
KEY `dt` (`dt`),
KEY `data` (`data`) USING BTREE,
KEY `test_udp_new_device_sn_dt_index` (`device_sn`,`dt`),
KEY `test_udp_new_device_sn_data_dt_index` (`device_sn`,`data`,`dt`)
) ENGINE=InnoDB AUTO_INCREMENT=44449751 DEFAULT CHARSET=latin1 ROW_FORMAT=DYNAMIC
The most frequent queries being run:
SELECT *
FROM test_udp_new
WHERE device_sn = 'xxx'
AND dt >= 'xxx'
AND dt <= 'xxx'
ORDER BY dt DESC;
The optimal way to handle that query is in a non-partitioned table with
INDEX(serial_number, dt)
Even better is to change the PRIMARY KEY. Assuming you currently have id AUTO_INCREMENT because there is not a unique combination of columns suitable for being a "natural PK",
PRIMARY KEY(serial_number, dt, id), -- to optimize that query
INDEX(id) -- to keep AUTO_INCREMENT happy
If there are other queries that are run often, please provide them; this may hurt them. In large tables, it is a juggling task to find the optimal index(es).
Other Comments:
There are very few use cases for which partitioning actually speed up processing.
Making lots of 'identical' tables is a maintenance nightmare, and, again, not a performance benefit. There are probably a hundred Q&A on stackoverflow shouting not to do such.
By having serial_number first in the PRIMARY KEY, all queries referring to a single serial_number are likely to benefit.
A million serial_numbers? No problem.
One common use case for partitioning involves purging "old" data. This is because big DELETEs are much more costly than DROP PARTITION. That involves PARTITION BY RANGE(TO_DAYS(dt)). If you are interested in that, my PK suggestion still stands. (And the query in question will run about the same speed with or without this partitioning.)
How many months before the table outgrows your disk? (If this will be an issue, let's discuss it.)
Do you need 8-byte DOUBLE? FLOAT has about 7 significant digits of precision and takes only 4 bytes.
You are using InnoDB?
Is serial_number fixed at 20 characters? If not, use VARCHAR. Also, CHARACTER SET ascii may be better than the default of utf8?
Each table (or each partition of a table) involves at least one file that the OS must deal with. When you have "too many", the OS groans, often before MySQL groans. (It is hard to make either "die" of overdose.)
Addressing the query
PRIMARY KEY (`id`),
KEY `device_sn_2` (`dt`,`device_sn`),
KEY `dt` (`dt`),
KEY `data` (`data`) USING BTREE,
KEY `test_udp_new_device_sn_dt_index` (`device_sn`,`dt`),
KEY `test_udp_new_device_sn_data_dt_index` (`device_sn`,`data`,`dt`)
-->
PRIMARY KEY(`device_sn`,`dt`, id),
INDEX(id)
KEY `dt_sn` (`dt`,`device_sn`),
KEY `data` (`data`) USING BTREE,
Notes:
By starting the PK with device_sn, dt, you get the clustering benefits to make the query with WHERE device_sn = .. AND dt BETWEEN ...
INDEX(id) is to keep AUTO_INCREMENT happy.
When you have INDEX(a,b), INDEX(a) is redundant.
The (20) is meaningless; id will max out at about 4 billion.
I tossed the last index because it is probably helped enough by the new PK.
lng decimal(10,5) -- Don't need 5 decimal places to left of point; only need 3 or 2. So: lat decimal(7,5),lng decimal(8,5)`. This will save a total of 3 bytes per row.
I have a monitoring table with the following structure:
CREATE TABLE `monitor_data` (
`monitor_id` INT(10) UNSIGNED NOT NULL,
`monitor_data_time` INT(10) UNSIGNED NOT NULL,
`monitor_data_value` INT(10) NULL DEFAULT NULL,
INDEX `monitor_id_data_time` (`monitor_id`, `monitor_data_time`),
INDEX `monitor_data_time` (`monitor_data_time`)
)
COLLATE='utf8_general_ci'
ENGINE=InnoDB;
This is a very high traffic table with potentially thousands of rows every minute. Each row belongs to a monitor and contains a value and time (unix_timestamp)
I have three issues:
1.
Suddenly, after a number of months in dev, the table suddenly became very slow. Queries that previously was done under a second could now take up to a minute. I'm using standard settings in my.cnf since this is a dev machine, but the behavior was indeed very strange to me.
2.
I'm not sure that I have optimal indexes. A "normal" query looks like this:
SELECT DISTINCT(md.monitor_data_time), monitor_data_value
FROM monitor_data md
WHERE md.monitor_id = 165
AND md.monitor_data_time >= 1484076760
AND md.monitor_data_time <= 1487271199
ORDER BY md.monitor_data_time ASC;
A EXPLAIN on the query above looks like this:
id;select_type;table;type;possible_keys;key;key_len;ref;rows;Extra
1;SIMPLE;md;range;monitor_id_data_time,monitor_data_time;monitor_id_data_time;8;\N;149799;Using index condition; Using temporary; Using filesort
What do you think about the indexes?
3.
If I leave out the DISTINCT in the query above, I actually get duplicate rows even though there aren't any duplicate rows in the table. Any explanation to this behavior?
Any input is greatly appreciated!
UPDATE 1:
New suggestion on table structure:
CREATE TABLE `monitor_data_test` (
`monitor_id` INT UNSIGNED NOT NULL,
`monitor_data_time` INT UNSIGNED NOT NULL,
`monitor_data_value` INT UNSIGNED NULL DEFAULT NULL,
PRIMARY KEY (`monitor_data_time`, `monitor_id`),
INDEX `monitor_data_time` (`monitor_data_time`)
) COLLATE='utf8_general_ci' ENGINE=InnoDB;
SELECT DISTINCT(md.monitor_data_time), monitor_data_value
is the same as
SELECT DISTINCT md.monitor_data_time, monitor_data_value
That is, the pair is distinct. It does not dedup just the time. Is that what you want?
If you are trying to de-dup just the time, then do something like
SELECT time, AVG(value)
...
GROUP BY time;
For optimal performance of
WHERE md.monitor_id = 165
AND md.monitor_data_time >= 14840767604 ...
you need
PRIMARY KEY (monitor_id, monitor_data_time)
and it must be in that order. The opposite order is much less useful. The guiding principle is: Start with the '=', then move on to the 'range'. More discussion here.
Do you have 4 billion monitor_id values? INT takes 4 bytes; consider using a smaller datatype.
Do you have other queries that need optimizing? It is better to design the index(es) after gather all the important queries.
Why PK
In InnoDB, the PRIMARY KEY is "clustered" with the data. That is, the data is an ordered list of triples: (id, time, value) stored in a B+Tree. Locating id = 165 AND time = 1484076760 is a basic operation of a BTree. And it is very fast. Then scanning forward (that's the "+" part of "B+Tree") until time = 1487271199 is a very fast operation of "next row" in this ordered list. Furthermore, since value is right there with the id and time, there is no extra effort to get the values.
You can't scan the requested rows any faster. But it requires PRIMARY KEY. (OK, UNIQUE(id, time) would be 'promoted' to be the PK, but let's not confuse the issue.)
Contrast... Given an index (time, id), it would do the scan over the dates fine, but it would have to skip over any entries where id != 165 But it would have to read all those rows to discover they do not apply. A lot more effort.
Since it is unclear what you intended by DISTINCT, I can't continue this detailed discussion of how that plays out. Suffice it to say: The possible rows have been found; now some kind of secondary pass is needed to do the DISTINCT. (It may not even need to do a sort.)
What do you think about the indexes?
The index on (monitor_id,monitor_data_time) seems appropriate for the query. That's suited to an index range scan operation, very quickly eliminating boatloads of rows that need to be examined.
Better would be a covering index that also includes the monitor_data_value column. Then the query could be satisfied entirely from the index, without a need to lookup pages from the data table to get monitor_data_value.
And even better would be having the InnoDB cluster key be the PRIMARY KEY or UNIQUE KEY on the columns, rather than incurring the overhead of the synthetic row identifier that InnoDB creates when an appropriate index isn't defined.
If I wasn't allowing duplicate (monitor_id, monitor_data_time) tuples, then I'd define the table with a UNIQUE index on those non-nullable columns.
CREATE TABLE `monitor_data`
( `monitor_id` INT(10) UNSIGNED NOT NULL
, `monitor_data_time` INT(10) UNSIGNED NOT NULL
, `monitor_data_value` INT(10) NULL DEFAULT NULL
, UNIQUE KEY `monitor_id_data_time` (`monitor_id`, `monitor_data_time`)
) ENGINE=InnoDB
or equivalent, specify PRIMARY in place of UNIQUE and remove the identifier
CREATE TABLE `monitor_data`
( `monitor_id` INT(10) UNSIGNED NOT NULL
, `monitor_data_time` INT(10) UNSIGNED NOT NULL
, `monitor_data_value` INT(10) NULL DEFAULT NULL
, PRIMARY KEY (`monitor_id`, `monitor_data_time`)
) ENGINE=InnoDB
Any explanation to this behavior?
If the query (shown in the question) returns a different number of rows with the DISTINCT keyword, then there must be duplicate (monitor_id,monitor_data_time,monitor_data_value) tuples in the table. There's nothing in the table definition that guarantees us that there aren't duplicates.
There are a couple of other possible explanations, but those explanations are all related to rows being added/changed/removed, and the queries seeing different snapshots, transaction isolation levels, yada, yada. If the data isn't changing, then there are duplicate rows.
A PRIMARY KEY constraint (or UNIQUE KEY constraint non-nullable columns) would guarantee us uniqueness.
Note that DISTINCT is a keyword in the SELECT list. It's not a function. The DISTINCT keyword applies to all expressions in the SELECT list. The parens around md.monitor_date_time are superfluous.
Leaving the DISTINCT keyword out would eliminate the need for the "Using filesort" operation. And that can be expensive for large sets, particularly when the set is too large to sort in memory, and the sort has to spill to disk.
It would be much more efficient to have guaranteed uniqueness, omit the DISTINCT keyword, and return rows in order by the index, preferably the cluster key.
Also, the secondary index monitor_data_time doesn't benefit this query. (There may be other queries that can make effective use of the index, though one suspects that those queries would also make effective use of a composite index that had monitor_data_time as the leading column.
I have a fairly simple query that is taking about 14 seconds to complete and I would like to speed it up. I think I have the correct indexes in place, but I'm not sure...
Here is the query
SELECT *
FROM opportunities
WHERE cid = 7785
AND STATUS != 4
AND otype != 200
AND links > 0
AND ontopic != 'F'
ORDER BY links DESC
LIMIT 0, 100;
Here is the table schema
CREATE TABLE `opportunities` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`cid` int(11) NOT NULL,
`url` varchar(900) CHARACTER SET utf8 NOT NULL,
`status` tinyint(4) NOT NULL,
`links` int(11) NOT NULL,
`otype` int(11) NOT NULL,
`reserved` tinyint(4) NOT NULL,
`ontopic` varchar(3) CHARACTER SET utf8 NOT NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `cid` (`cid`,`url`),
KEY `cid1` (`cid`),
KEY `url` (`url`),
KEY `otype` (`otype`),
KEY `reserved` (`reserved`),
KEY `ontopic` (`ontopic`),
KEY `status` (`status`),
KEY `links` (`links`),
KEY `ontopic_links` (`ontopic`,`links`),
KEY `cid_status_otype_links_ontopic` (`cid`,`status`,`otype`,`links`,`ontopic`)
) ENGINE=InnoDB AUTO_INCREMENT=13022832 DEFAULT CHARSET=latin1
Here is the result of the EXPLAIN command
id: 1
select_type: Simple
table: opportunities
partitions: null
type: range
possible_keys: cid,cid1,otype,ontopic,status,links,ontopic_links,cid_status_otype_links_ontopic
key: links
keylen: 4
ref: null
rows: 1531552
filtered: 0.33
Extra: Using index condition; Using where
Thoughts / Questions
Am I reading it correctly that it is using the "links" key to do the query? Why wouldn't it use a more complete index, like the cid_status_otype_links_ontopic which covers all the conditions of my query?
Thanks in advance!
As requested
There are 30,961 results that match the query when you remove the LIMIT 0,100. Interestingly, the "count()" command returns almost instantaneously.
It's a funny thing about using inequality comparisons, that they count as range conditions.
That is, equality matches one value, but anything other than equality (!=, >, <, IN, BETWEEN).
By matching multiple values, it means that only the first column in an index used in a range condition is going to be optimized. You'd think that your index cid_status_otype_links_ontopic has all the columns mentioned in conditions of your query, but only the first two will be used. The first because you have an equality comparison for cid. The second because the next column is used in an inequality comparison, and then that's where it stops using columns from the index.*
Evidence: if you can force that index to be used, you should see the keylen field of the EXPLAIN result show only 5, which is the size of cid (4 bytes) + status (1 byte).
The MySQL optimizer apparently has predicted that it would be more beneficial to use your links index, because that allows it to access the rows in index order, which is the same as the sort order you requested with your ORDER BY.
Evidence: you don't see "Using filesort" in your EXPLAIN notes.
Is that really better than using one of the other indexes? Maybe, maybe not. The optimizer's predictions aren't always perfect.
You can use an index hint to override the optimizer's choice:
SELECT * FROM opportunities USE INDEX (cid_status_otype_links_ontopic) WHERE ...
Try that out, do the EXPLAIN of that query and compare it to your other EXPLAIN. Then execute both queries and see which is reliably faster.
(* Actually, I have to add a footnote about the index column usage. MySQL 5.6 and later can do a little bit better than just the two columns, when you see the note "Using Index Condition" in the EXPLAIN. But it's not quite the same. You can read more about that here: https://dev.mysql.com/doc/refman/5.6/en/index-condition-pushdown-optimization.html)
What you have must plow through all of the rows, using your 5-column index, then sort the results and deliver 100 rows.
The only index likely to be useful is INDEX(cid, links). This is because cid is the only column being tested with =, then having links might be useful for the ORDER BY and LIMIT. There is still the risk that the != tests will require filtering a lot of rows.
Are status and otype multi-valued? If either has only 2 values, then turning the != into = and adding it to the index would be beneficial.
Do you really need all the columns (SELECT *)? If not, and if you don't need any big columns (url), then you could go with a 'covering' index.
More on writing indexes .
Ok, I have the following MySQL table structure:
CREATE TABLE `creditlog` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`memberId` int(10) unsigned NOT NULL,
`quantity` decimal(10,2) unsigned DEFAULT NULL,
`timeAdded` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
`reference` varchar(255) DEFAULT NULL,
PRIMARY KEY (`id`),
KEY `memberId` (`memberId`),
KEY `timeAdded` (`timeAdded`));
And I'm querying it like this:
SELECT SUM(quantity) FROM creditlog where timeAdded>'2016-09-01' AND timeAdded<'2016-10-01' AND memberId IN (3,6,8,9,11)
Now, I also use the use index (timeAdded) because due to the number of entries it is more convenient. Explaining the above query shows:
type -> range,
key -> timeAdded,
rows -> 921294
extra -> using where
Meanwhile if I use the memberId INDEX it shows:
type -> range,
key -> memberId,
rows -> 1707849
extra -> using where
Now, my question is it's possible to combine these 2 indexes somehow to be used together and reduce the surface of the query since I ll also need to add more conditions (on other columns).
MySQL almost never uses two indexes in a single query; it is just not cost effective. However, composite indexes are often very efficient. You need this order: INDEX(memberId, timeAdded).
Build the index this way...
First include column(s) that are in the WHERE clause tested with =. (None, in your case.)
Any column(s) with IN.
One 'range', such as <, BETWEEN, etc.
Move onto all the fields of the GROUP BY or ORDER BY. (Not relevant here.)
There are a lot of exceptions and caveats. Some are given in my cookbook .
(Contrary to popular opinion, cardinality is almost never relevant in designing an index.)
Here is a way to compare two indexes (even with a table that is too small to get reliable timings):
FLUSH STATUS;
SELECT SQL_NO_CACHE ...;
SHOW SESSION STATUS LIKE 'Handler%';
(repeat for other query/index)
Smaller numbers almost always indicate better.
"timeAdded>'2016-09-01' AND timeAdded<'2016-10-01'" -- That excludes midnight on the first day. I recommend this pattern:
timeAdded >= '2016-09-01'
AND timeAdded < '2016-09-01' + INTERVAL 1 MONTH
That also avoids computing dates.
That smells like a common query? Have you considered building and maintaining Summary tables ? The equivalent query would probably run 10 times as fast.
I'm learning MySQL index and found that index should be applied to any column named in the WHERE clause of a SELECT query.
Then I found Multiple Column Index vs Multiple Indexes.
First Q, I was wondering what is multiple column index. I found code bellow from Joomla, is this Multiple Column Index?
CREATE TABLE `extensions` (
`extension_id` INT(11) NOT NULL AUTO_INCREMENT,
`name` VARCHAR(100) NOT NULL,
`type` VARCHAR(20) NOT NULL,
`element` VARCHAR(100) NOT NULL,
`folder` VARCHAR(100) NOT NULL,
`client_id` TINYINT(3) NOT NULL,
... ...
PRIMARY KEY (`extension_id`),
// does code below is multiple column index?
INDEX `element_clientid` (`element`, `client_id`),
INDEX `element_folder_clientid` (`element`, `folder`, `client_id`),
INDEX `extension` (`type`, `element`, `folder`, `client_id`)
)
Second Q, am I correct if thinking that one Multiple Column Index is used on one SELECT ?
SELECT column_x WHERE element=y AND clinet_id=y; // index: element_clientid
SELECT ex.col_a, tb.col_b
FROM extensions ex
LEFT JOIN table2 tb
ON (ex.ext_id = tb.ext_id)
WHERE ex.element=x AND ex.folder=y AND ex.client_id=z; // index: element_folder_clientid
General rule of thumb for indexes is to slap one onto any field used in a WHERE or JOIN clause.
That being said, there are some optimizations you can do. If you KNOW that a certain combination of fields are the only one that will ever be used in WHERE on a particular table, then you can create a single multi-field key on just those fields, e.g.
INDEX (field1, field2, field5)
v.s.
INDEX (field1),
INDEX (field2),
INDEX (field5)
A multi-field index can be more efficient in many cases, v.s having to scan multiple indexes. The downside is that the multi-field index is only usable if the fields in question are actually used in a WHERE clause.
With your sample queries, since element and field_id are in all three indexes, you might be better off splitting them off into their own dedicated index. If these are changeable fields, then it's better to keep it their own dedicated index. e.g. if you ever have to change field_id in bulk, the DB has to update 3 different indexes, v.s. updating just one dedicated one.
But it all comes down to benchmarking - test your particular setup with various index setups and see which performs best. Rules of thumbs are handy, but don't work 100% of the time.