Strange behaviour on MYSQL querying huge table - mysql

Question

I'm trying to understand a strange performance behaviour happening in a MYSQL data structure that I'm working on:
CREATE TABLE metric_values
(
dmm_id INT NOT NULL,
dtt_id BIGINT NOT NULL,
cus_id INT NOT NULL,
nod_id INT NOT NULL,
dca_id INT NULL,
value DOUBLE NOT NULL
)
ENGINE = InnoDB;
CREATE INDEX metric_values_dmm_id_index
ON metric_values (dmm_id);
CREATE INDEX metric_values_dtt_index
ON metric_values (dtt_id);
CREATE INDEX metric_values_cus_id_index
ON metric_values (cus_id);
CREATE INDEX metric_values_nod_id_index
ON metric_values (nod_id);
CREATE INDEX metric_values_dca_id_index
ON metric_values (dca_id);
CREATE TABLE dim_metric
(
dmm_id INT AUTO_INCREMENT
PRIMARY KEY,
met_id INT NOT NULL,
name VARCHAR(45) NOT NULL,
instance VARCHAR(45) NULL,
active BIT DEFAULT b'0' NOT NULL
)
ENGINE = InnoDB;
CREATE INDEX dim_metric_dmm_id_met_id_index
ON dim_metric (dmm_id, met_id);
CREATE INDEX dim_metric_met_id_index
ON dim_metric (met_id);
CONTEXT:
Some context, I'm trying to understand some strange performance behaviour happening in a data structure that I'm working on:
CREATE TABLE metric_values
(
dmm_id INT NOT NULL,
dtt_id BIGINT NOT NULL,
cus_id INT NOT NULL,
nod_id INT NOT NULL,
dca_id INT NULL,
value DOUBLE NOT NULL
)
ENGINE = InnoDB;
CREATE INDEX metric_values_dmm_id_index
ON metric_values (dmm_id);
CREATE INDEX metric_values_dtt_index
ON metric_values (dtt_id);
CREATE INDEX metric_values_cus_id_index
ON metric_values (cus_id);
CREATE INDEX metric_values_nod_id_index
ON metric_values (nod_id);
CREATE INDEX metric_values_dca_id_index
ON metric_values (dca_id);
CREATE TABLE dim_metric
(
dmm_id INT AUTO_INCREMENT
PRIMARY KEY,
met_id INT NOT NULL,
name VARCHAR(45) NOT NULL,
instance VARCHAR(45) NULL,
active BIT DEFAULT b'0' NOT NULL
)
ENGINE = InnoDB;
CREATE INDEX dim_metric_dmm_id_met_id_index
ON dim_metric (dmm_id, met_id);
CREATE INDEX dim_metric_met_id_index
ON dim_metric (met_id);
CONTEXT:
Metric_values have something close to 100 milion rows and table dim_metric has 1024 rows.
I'm doing a simple JOIN between this 2 tables and I'm having huge performance issues. Trying to figure out what the problem is I stumbled in this strange behaviour.
I can't execute the JOIN using the column met_id as a filter. I left it running for 10 minutes and lost the connection to the database due to timeout before I got any results back;
Running a explain on the query I can see that the indexes are being used correctly (I assume) and only 1052 rows are being scanned from metric_values.
EXPLAIN
SELECT
count(0)
FROM metric_values v
INNER JOIN dim_metric m ON m.dmm_id = v.dmm_id
WHERE 1=1
AND m.met_id = 1;
1 SIMPLE m ref PRIMARY,dim_metric_met_id_index,dim_metric_dmm_id_met_id_index dim_metric_met_id_index 4 const 1 Using index
1 SIMPLE v ref metric_values_dmm_id_index metric_values_dmm_id_index 4 oi_fact.m.dmm_id 1052 Using index
Doing a simple change to the query to use a sub select instead of a JOIN I can get the results after ~45 seconds.
Running a explain on the modified query I can see that the index is not the primary resource being used to fetch the data and that almost 20 million rows were scaned to bring me the result.
EXPLAIN
SELECT
count(0)
FROM metric_values v
WHERE 1=1
AND v.dmm_id = (SELECT m.dmm_id FROM dim_metric m WHERE m.met_id = 1);
1 PRIMARY v ref metrics_values_dmm_id_index metrics_values_dmm_id_index 4 const 19589800 Using where; Using index
2 SUBQUERY m ref dim_metric_met_id_index dim_metric_met_id_index 4 const 1 Using index
Can someone explain to me what is happening? Did I misunderstood what the EXPLAIN is telling me? Can I do some changes to the data model to improve the query performance?