💾 Archived View for thebird.nl › gn-gemtext-threads › issues › systems › mariadb › ProbeSetXRef.gmi captured on 2023-05-24 at 18:22:15. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2022-07-16)
-=-=-=-=-=-=-
Juggling indexes and transforming to InnoDB led to a massive speed increase for typical ProbeSetXRef queries. Global search for brca2 went down from 13s to 4s! Try
http://genenetwork.org/gsearch?type=gene&terms=brca2
Not only global search is faster, *all* search is significantly faster.
File data size increased from 3.3GB to 5.4GB (1.6x).
Note that the full wildcard '*' search throws an error or empty results. That needs a fix.
A primary key index was introduced (required by innodb) and joined indexes were split in two. I also reduced the text index to 5 characters for each locus to reduce the data pump.
It turned out that a join on a mixed latin1 and utf8 field does not work which makes sense when you think about it. Here ProbeSetXRef.locus joins against geno.name.
The short of it is that it is ill advised to change the charset table by table! So, we'll stick with latin1 until we convert all tables. Needs proper testing anyway.
Some optimisation is still possible - removing the Old columns and perhaps the Locus VARCHAR column can be reduced in size.
Using prototocol from
SHOW CREATE TABLE ProbeSetXRef;
| ProbeSetXRef | CREATE TABLE 'ProbeSetXRef' (
'ProbeSetFreezeId' smallint(5) unsigned NOT NULL DEFAULT 0,
'ProbeSetId' int(10) unsigned NOT NULL DEFAULT 0,
'DataId' int(10) unsigned NOT NULL DEFAULT 0,
'Locus_old' char(20) DEFAULT NULL,
'LRS_old' double DEFAULT NULL,
'pValue_old' double DEFAULT NULL,
'mean' double DEFAULT NULL,
'se' double DEFAULT NULL,
'Locus' varchar(50) DEFAULT NULL,
'LRS' double DEFAULT NULL,
'pValue' double DEFAULT NULL,
'additive' double DEFAULT NULL,
'h2' float DEFAULT NULL,
UNIQUE KEY 'ProbeSetId' ('ProbeSetFreezeId','ProbeSetId'),
UNIQUE KEY 'DataId_IDX' ('DataId'),
KEY 'Locus_IDX' ('Locus')
) ENGINE=MyISAM DEFAULT CHARSET=latin1 |
For every probe set (read dataset measuring point):
select * from ProbeSetXRef limit 2; | ProbeSetFreezeId | ProbeSetId | DataId | Locus_old | LRS_old | pValue_old | mean | se | Locus | LRS | pValue | additive | h2 | +------------------+------------+--------+------------+--------------------+------------+------------------+---------------------+------------+------------------+--------+--------------------+------+ | 1 | 1 | 1 | 10.095.400 | 13.3971627898894 | 0.163 | 5.48794285714286 | 0.08525787814808819 | rs13480619 | 12.590069931048 | 0.269 | -0.28515625 | NULL | | 1 | 2 | 2 | D15Mit189 | 10.042057464356201 | 0.431 | 9.90165714285714 | 0.0374686634976217 | rs29535974 | 10.5970737900941 | 0.304 | -0.116783333333333 | NULL | +------------------+------------+--------+------------+--------------------+------------+------------------+---------------------+------------+------------------+--------+--------------------+------+
where ProbeSetFreezeId is the dataset (experiment). ProbesetId refers
to the probe set information (measuring point). DataId points to the
data point. The other values are used for search.
MariaDB [db_webqtl]> select count(*) from ProbeSetXRef; +----------+ | count(*) | +----------+ | 48076905 | +----------+
rg ProbeSetXRef --color=always |less -R
rg ProbeSetXRef --type=py -l|fzf --preview="rg --color=always -A 20 ProbeSetXRef {}" --preview-window=right:85%:wrap
In SQL where statements these are usually combined
ProbeSetXRef.ProbeSetFreezeId
ProbeSetXRef.ProbeSetId
and sometimes
ProbeSetXRef.DataId
ProbeSetXRef.Locus
As can be seen from above table definition the indices are matching
UNIQUE KEY 'ProbeSetId' ('ProbeSetFreezeId','ProbeSetId'),
UNIQUE KEY 'DataId_IDX' ('DataId'),
KEY 'Locus_IDX' ('Locus')
The combination of
select count(distinct ProbeSetFreezeId,ProbeSetId) from ProbeSetXRef limit 10;
is unique. Now we should also notice that DataId is unique and will make a smaller primary index. It is wort trying to split the combined ('ProbeSetFreezeId','ProbeSetId') into two indices. I'll do that once we are on innodb. Oh wait, I have to set the primary key first, I don't think I can change that. Let's create a new table for testing:
CREATE TABLE mytest AS SELECT * FROM ProbeSetXRef; ALTER TABLE mytest ADD PRIMARY KEY(DataId); ALTER TABLE mytest ADD KEY(ProbeSetFreezeId); ALTER TABLE mytest ADD KEY(ProbeSetId); SHOW CREATE TABLE mytest;
Some select statement and maybe a page of GN2.
MariaDB [db_webqtl]> select count(*) from ProbeSetXRef where ProbeSetFreezeId<200 and ProbeSetId<1000 and pValue>0.5; +----------+ | count(*) | +----------+ | 19068 | +----------+ 1 row in set (1.752 sec) MariaDB [db_webqtl]> select count(*) from ProbeSetXRef where ProbeSetFreezeId<300 and ProbeSetId<1000 and pValue>0.5; +----------+ | count(*) | +----------+ | 19068 | +----------+ 1 row in set (13.781 sec)
And InnoDB mytest has a 3x speedup for the real query:
MariaDB [db_webqtl]> select count(*) from ProbeSetXRef where ProbeSetFreezeId<200 and ProbeSetId<1000 and pValue>0.5; +----------+ | count(*) | +----------+ | 19068 | +----------+ 1 row in set (1.748 sec) MariaDB [db_webqtl]> select count(*) from mytest where ProbeSetFreezeId<200 and ProbeSetId<1000 and pValue>0.5; +----------+ | count(*) | +----------+ | 19068 | +----------+ 1 row in set (0.445 sec)
Note the second query traverses the full file and the mytest version is the same speed.
ALTER TABLE ProbeSetXRef ADD PRIMARY KEY(ProbeSetFreezeId,ProbeSetId);
MariaDB [db_webqtl]> DROP INDEX ProbeSetId ON ProbeSetXRef; MariaDB [db_webqtl]> show index from ProbeSetXRef;
ALTER TABLE ProbeSetXRef ENGINE = InnoDB;
There are some text fields for locus, I think it is safe to translate those to utf8.
ALTER TABLE mytest CONVERT TO CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci; ALTER TABLE mytest ADD KEY(Locus(5));
Note I specified a size for Locus to keep the index nimble.
For those cases see bottom of move-to-innodb.gmi.
OPTIMIZE NO_WRITE_TO_BINLOG TABLE mytest;
RENAME TABLE ProbeSetXRef TO ProbeSetXRef_old, mytest TO ProbeSetXRef;
| mytest | CREATE TABLE 'mytest' (
'ProbeSetFreezeId' smallint(5) unsigned NOT NULL DEFAULT 0,
'ProbeSetId' int(10) unsigned NOT NULL DEFAULT 0,
'DataId' int(10) unsigned NOT NULL DEFAULT 0,
'Locus_old' char(20) DEFAULT NULL,
'LRS_old' double DEFAULT NULL,
'pValue_old' double DEFAULT NULL,
'mean' double DEFAULT NULL,
'se' double DEFAULT NULL,
'Locus' varchar(50) DEFAULT NULL,
'LRS' double DEFAULT NULL,
'pValue' double DEFAULT NULL,
'additive' double DEFAULT NULL,
'h2' float DEFAULT NULL,
PRIMARY KEY ('DataId'),
KEY 'ProbeSetFreezeId' ('ProbeSetFreezeId'),
KEY 'ProbeSetId' ('ProbeSetId'),
KEY 'Locus_2' ('Locus'(5))
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 |
MariaDB [db_webqtl]> select count(*) from ProbeSetXRef where ProbeSetFreezeId<200 and ProbeSetId<1000 and pValue>0.5; +----------+ | count(*) | +----------+ | 19068 | +----------+ 1 row in set (0.058 sec)
Now I had 3 pages open
http://genenetwork.org/show_trait?trait_id=1427571_at&dataset=HC_M2_0606_P
http://genenetwork.org/gsearch?type=gene&terms=brca2
The first two pages loaded the same. But the third (global search) took much longer than the 15 seconds it did before I modified ProbeSetXRef. Doh! It should be
It took the server 15.71517s seconds to process this page. It took your browser 0.598 second(s) to render this page
The query timed out, so I could look in the slow query log we have.
(But first I added a standard test on sheepdog to make sure we see these regressions.)
vim /var/log/mysql/mysql-slow.log
# Time: 211230 8:20:36
# User@Host: webqtlout[webqtlout] @ localhost []
# Thread_id: 11771 Schema: db_webqtl QC_hit: No
# Query_time: 470.965035 Lock_time: 0.000258 Rows_sent: 1017 Rows_examined: 115097830
# Rows_affected: 0 Bytes_sent: 253724
SELECT ProbeSetFreeze.'Name', ProbeSetFreeze.'FullName', ProbeSet.'Name', ProbeSet.'Symbol', CAST(ProbeSet.'description' AS BINARY), CAST(ProbeSet.'Probe_Target_Description' AS BINARY), ProbeSet.'Chr', ProbeSet.'Mb', ProbeSetXRef.'Mean', ProbeSetXRef.'LRS', ProbeSetXRef.'Locus', ProbeSetXRef.'pValue', ProbeSetXRef.'additive', Geno.'Chr' as geno_chr, Geno.'Mb' as geno_mb FROM Species
INNER JOIN InbredSet ON InbredSet.'SpeciesId'= Species.'Id'
INNER JOIN ProbeFreeze ON ProbeFreeze.'InbredSetId' = InbredSet.'Id' INNER JOIN Tissue ON ProbeFreeze.'TissueId' = Tissue.'Id'
INNER JOIN ProbeSetFreeze ON ProbeSetFreeze.'ProbeFreezeId' = ProbeFreeze.'Id'
INNER JOIN ProbeSetXRef ON ProbeSetXRef.'ProbeSetFreezeId' = ProbeSetFreeze.'Id'
INNER JOIN ProbeSet ON ProbeSet.'Id' = ProbeSetXRef.'ProbeSetId'
LEFT JOIN Geno ON ProbeSetXRef.'Locus' = Geno.'Name' AND Geno.'SpeciesId' = Species.'Id'
WHERE ((((MATCH (ProbeSet.Name, ProbeSet.description, ProbeSet.symbol, alias, GenbankId, UniGeneId, Probe_Target_Description) AGAINST ('sh*' IN BOOLEAN MODE))) AND ProbeSet.Id = ProbeSetXRef.ProbeSetId and ProbeSetXRef.ProbeSetFreezeId = 112 )) and ProbeSet.Id = ProbeSetXRef.ProbeSetId and ProbeSetXRef.ProbeSetFreezeId = 112 ORDER BY ProbeSet.symbol ASC;
this is a different query which logs after a >60s query (it is one of my monitors looking for 'sh*'. This is an interesting one to look at, but it does not match my query.
The actual query showed up as
SELECT
Species.'Name' AS species_name,
InbredSet.'Name' AS inbredset_name,
Tissue.'Name' AS tissue_name,
ProbeSetFreeze.Name AS probesetfreeze_name,
ProbeSetFreeze.FullName AS probesetfreeze_fullname,
ProbeSet.Name AS probeset_name,
ProbeSet.Symbol AS probeset_symbol,
CAST(ProbeSet.'description' AS BINARY) AS probeset_description,
ProbeSet.Chr AS chr,
ProbeSet.Mb AS mb,
ProbeSetXRef.Mean AS mean,
ProbeSetXRef.LRS AS lrs,
ProbeSetXRef.'Locus' AS locus,
ProbeSetXRef.'pValue' AS pvalue,
ProbeSetXRef.'additive' AS additive,
ProbeSetFreeze.Id AS probesetfreeze_id,
Geno.Chr as geno_chr,
Geno.Mb as geno_mb
FROM Species
INNER JOIN InbredSet ON InbredSet.'SpeciesId'=Species.'Id'
INNER JOIN ProbeFreeze ON ProbeFreeze.InbredSetId=InbredSet.'Id'
INNER JOIN Tissue ON ProbeFreeze.'TissueId'=Tissue.'Id'
INNER JOIN ProbeSetFreeze ON ProbeSetFreeze.ProbeFreezeId=ProbeFreeze.Id
INNER JOIN ProbeSetXRef ON ProbeSetXRef.ProbeSetFreezeId=ProbeSetFreeze.Id
INNER JOIN ProbeSet ON ProbeSet.Id = ProbeSetXRef.ProbeSetId
LEFT JOIN Geno ON ProbeSetXRef.Locus = Geno.Name AND Geno.SpeciesId = Species.Id
WHERE ( MATCH (ProbeSet.Name,ProbeSet.description,ProbeSet.symbol,ProbeSet.alias,ProbeSet.GenbankId, ProbeSet.UniGeneId, ProbeSet.Probe_Target_Description) AGAINST ('brca2' IN BOOLEAN MODE) )
AND ProbeSetFreeze.confidentiality < 1
AND ProbeSetFreeze.public > 0
ORDER BY species_name, inbredset_name, tissue_name, probesetfreeze_name, probeset_name
LIMIT 6000;
which looks somewhat similar. There are some joins and I think my index size restriction of Locus(5) may be the problem here. The INNER JOIN keyword selects records that have matching values in both tables, so maybe the Locus has trouble matching.
Add a full index:
ALTER TABLE ProbeSetXRef ADD KEY(Locus);
The alternative is that mixed innodb and myisam joins are expensive, or more likely the mixed charsets are not playing well. I can try the EXPLAIN statement next.
Well, it turned out that a join on a mixed latin1 and utf8 field does not work which makes sense when you think about it. It joins against geno.name. Now in GN geno.name is usally compared against a string or ProbeSetXRef.locus.
The short of it is that it is ill advised to change the charset table by table! So, we'll stick with latin1 until we convert all tables.
SELECT
Species.'Name' AS species_name,
InbredSet.'Name' AS inbredset_name,
Tissue.'Name' AS tissue_name,
ProbeSetFreeze.Name AS probesetfreeze_name,
ProbeSetFreeze.FullName AS probesetfreeze_fullname,
ProbeSet.Name AS probeset_name,
ProbeSet.Symbol AS probeset_symbol,
CAST(ProbeSet.'description' AS BINARY) AS probeset_description,
ProbeSet.Chr AS chr,
ProbeSet.Mb AS mb,
mytest.Mean AS mean,
mytest.LRS AS lrs,
mytest.'Locus' AS locus,
mytest.'pValue' AS pvalue,
mytest.'additive' AS additive,
ProbeSetFreeze.Id AS probesetfreeze_id,
Geno.Chr as geno_chr,
Geno.Mb as geno_mb
FROM Species
INNER JOIN InbredSet ON InbredSet.'SpeciesId'=Species.'Id'
INNER JOIN ProbeFreeze ON ProbeFreeze.InbredSetId=InbredSet.'Id'
INNER JOIN Tissue ON ProbeFreeze.'TissueId'=Tissue.'Id'
INNER JOIN ProbeSetFreeze ON ProbeSetFreeze.ProbeFreezeId=ProbeFreeze.Id
INNER JOIN mytest ON mytest.ProbeSetFreezeId=ProbeSetFreeze.Id
INNER JOIN ProbeSet ON ProbeSet.Id = mytest.ProbeSetId
LEFT JOIN Geno ON mytest.Locus = Geno.Name AND Geno.SpeciesId = Species.Id
WHERE ( MATCH (ProbeSet.Name,ProbeSet.description,ProbeSet.symbol,ProbeSet.alias,ProbeSet.GenbankId, ProbeSet.UniGeneId, ProbeSet.Probe_Target_Description) AGAINST ("brca2" IN BOOLEAN MODE) )
AND ProbeSetFreeze.confidentiality < 1
AND ProbeSetFreeze.public > 0
ORDER BY species_name, inbredset_name, tissue_name, probesetfreeze_name, probeset_name LIMIT 600;
Should run 600 rows in set (11.638 sec)
Changing the index now does 6000 rows in set (0.240 sec)!
The new table structure is
| ProbeSetXRef | CREATE TABLE 'ProbeSetXRef' (
'ProbeSetFreezeId' smallint(5) unsigned NOT NULL DEFAULT 0,
'ProbeSetId' int(10) unsigned NOT NULL DEFAULT 0,
'DataId' int(10) unsigned NOT NULL DEFAULT 0,
'Locus_old' char(20) DEFAULT NULL,
'LRS_old' double DEFAULT NULL,
'pValue_old' double DEFAULT NULL,
'mean' double DEFAULT NULL,
'se' double DEFAULT NULL, 'Locus' varchar(50) DEFAULT NULL, 'LRS' double DEFAULT NULL,
'pValue' double DEFAULT NULL,
'additive' double DEFAULT NULL, 'h2' float DEFAULT NULL,
PRIMARY KEY ('DataId'),
KEY 'ProbeSetFreezeId' ('ProbeSetFreezeId'),
KEY 'ProbeSetId' ('ProbeSetId'),
KEY 'Locus_2' ('Locus'(5))
) ENGINE=InnoDB DEFAULT CHARSET=latin1 |