Can MySql nested SP be a bottleneck?

Question

We have this MySQL SP, which calls a nested SP. It seems it does NOT perform well under load.

It is possible that this SP becomes slow under load because it calls a nested SP and uses temporary tables to pass the data to main SP?

DELIMITER $$

drop procedure if exists `GeoAreaFlattened_Select`;

create procedure `GeoAreaFlattened_Select`(
    _areas MEDIUMTEXT,
    _comparisonGroup varchar(21844),
    _parentArea varchar(21844),
    _areaType varchar(21844)
)
begin

drop temporary table if exists areas;

-- areas
call CreateAreas(_areas, _comparisonGroup, _parentArea, _areaType);

SELECT
    areas.ID,
    areas.Code,
    areas.Name,
    areas.LevelId,
    GeoAreaLevel.Name AS AreaTypeLabel,
    GeoAreaLevel.Identifier AS AreaTypeIdentifier
FROM
    areas
INNER JOIN
    GeoAreaLevel
ON
    areas.levelid = GeoAreaLevel.id
ORDER BY areas.name ASC;

drop temporary table areas;

end

The nested SP:

-- --------------------------------------------------------------------------------
-- Routine DDL
-- Note: comments before and after the routine body will not be stored by the server
-- --------------------------------------------------------------------------------
drop procedure if exists `CreateAreas`;

DELIMITER $$

CREATE PROCEDURE `CreateAreas`(
    _areas varchar(21844),
    _comparisonGroup varchar(21844),
    _parentArea varchar(21844),
    _areaType varchar(21844)
)
BEGIN

    -- create temporary table "areas"
    -- fill with area ids

    create temporary table areas (
        id int not null,
        code varchar(30),
        name varchar(100),
        shortName varchar(100),
        levelid int not null,
        sortOrder int not null,
        key (id)
    );

    -- assumes that only one of the 3 options is valid, areas, comparison group, bounded comparison group

    if (_areas is not null) then

        set @sql = concat('insert into areas (id, code, name, shortName, levelid, sortOrder) select id, Code, Name, ShortName, LevelID, 0 from GeoArea where Code in (''', replace(_areas, ',', ''','''), ''')');
        prepare stmt from @sql;
        execute stmt;
        deallocate prepare stmt;

    elseif (_comparisonGroup is not null) then

        -- might not be the most efficient way, but is consistent with the approach above, and we do not expect the list to be long
        insert into areas (id, code, name, shortName, levelid, sortOrder)
        select GeoAreaID, GeoArea.Code, GeoArea.Name, GeoArea.ShortName, GeoArea.LevelID, SortOrder
        from ComparisonGroupGeoAreaLink
        INNER JOIN
        GeoArea
        ON GeoArea.ID = GeoAreaID
        where ComparisonGroupID = (select id from ComparisonGroup where Identifier = _comparisonGroup)
        and IsMember = 1;

    elseif (_parentArea is not null and _areaType is not null) then

        -- might not be the most efficient way, but is consistent with the approach above, and we do not expect the list to be long
        insert into areas (id, code, name, shortName, levelid, sortOrder)
    select a.ID, a.Code, a.Name, a.ShortName, a.LevelID, 0
        from (select id from GeoArea where Code = _parentArea) as t
        INNER JOIN
        GeoAreaLinkCache c
        ON
        c.ParentAreaID = t.id
        inner join GeoArea a
        on c.ChildAreaID = a.ID
        INNER JOIN
        (select id from GeoAreaLevel where Identifier = _areaType) as l
        ON
        a.LevelID = l.id;        

    elseif (_areaType is not null) then

        -- might not be the most efficient way, but is consistent with the approach above, and we do not expect the list to be long
        set @sql = concat('insert into areas (id, code, name, shortName, levelid, sortOrder)
        select a.ID, a.Code, a.Name, a.ShortName, a.LevelID, 0
        from 
        (select id from GeoAreaLevel where Identifier in (''', replace(_areaType, ',', ''','''), ''')) l
        INNER JOIN
        GeoArea a
        ON
        a.LevelID = l.id');
        prepare stmt from @sql;
        execute stmt;
        deallocate prepare stmt;


    end if;                 

END

Answer 1

Yes, it's possible. We don't have measurements of how long this takes, but one can expect a temp table to cause some overhead as you create it, write data into it, query it, then drop it. How often is this being called?

Also, MySQL's stored procedures are generally known to be pretty inefficient. They don't retain the compiled form of the procedure, as they do in Oracle and other RDBMS brands. In other words, every session recompiles each procedure it uses the first time it is called.

First, I would suggest eliminating the temp table. Just design the nested procedure to build the right SQL query as a string, and return that string. Then the outer procedure executes the query is its result. You should be able to skip the create/drop of the temp table, and the insert into the temp table.

Second, if I were designing this app, I don't see any need for either stored procedure. I'd recommend writing code to build the right SQL query in your application, and then just execute that query from the application.

Answer 2

It seems nested SPs was NOT the main bottleneck.

I've changed temp tables to MEMORY engine and it really did the trick and made AMAZING difference. Solution was suggested in

https://dba.stackexchange.com/questions/52825/can-mysql-nested-sp-be-a-bottleneck/52863?noredirect=1#52863

create temporary table areas (
    id int not null,
    code varchar(30),
    name varchar(100),
    shortName varchar(100),
    levelid int not null,
    sortOrder int not null,
    key (id)
) ENGINE=MEMORY;