Выберите Оптимизация запросов на MySQL для больших объемов данных

Question

## MySql Server 5.5, двигатель базы данных MyISAM, таблица fact_transaction индексируется на date_key, time_key, unit_cost_price, unit_retail_price (комбинированный ключ) и то же самое для fact_stockout_sales за исключением time_key. ##

Query plan

Запрос

SELECT 
    t.Level, t.Name, t.KeyValue, 
    ROUND((SUM(t.Gross)/SUM(t.Revenue))*100, 2) AS Value, 
    ROUND((SUM(t.adjustedGross)/SUM(t.adjustedRevenue))*100, 2) AS adjustedValue, 
    t.dataType AS dataType 
FROM 
    (SELECT "item" AS Level, ds.product_name AS Name, ds.product_id AS KeyValue, 
     SUM(ft.gross_profit) AS Gross, 
     SUM(ft.selling_amount) AS Revenue, 
     SUM(ft.adjusted_gross_profit) AS adjustedGross, 
     SUM(ft.adjusted_selling_amount) AS adjustedRevenue, 
     "%" AS dataType 
    FROM fact_transaction AS ft 
    JOIN dim_sku AS ds ON ft.sku_key = ds.sku_key 
    WHERE ft.date_key BETWEEN 20080215 AND 20130107 
     AND ft.time_key BETWEEN 100 AND 235900 
     AND ft.unit_cost_price BETWEEN 0 AND 1333 
     AND ft.unit_retail_price BETWEEN 0 AND 16500 
     AND ft.store_key IN ("1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15", "16") 
     AND ds.product_id IN (1312009,1312007,... Huge List say 30000) 
     AND ds.category IN ("Male","Female","Unisex") 
     AND ft.day_of_week IN ("1","2","3","4","5","6","7") 
     AND ds.collection_name IN ("Base","SS12","AW12") 
    GROUP BY ds.product_id       
    UNION 
    SELECT "item" AS Level, ds.product_name AS Name, ds.product_id AS KeyValue, 
     SUM(ft.gross_profit) AS Gross, 
     SUM(ft.selling_amount) AS Revenue, 
     SUM(ft.adjusted_gross_profit) AS adjustedGross, 
     SUM(ft.adjusted_selling_amount) AS adjustedRevenue, 
     "%" AS dataType 
    FROM fact_stockout_sales AS ft 
    JOIN dim_sku AS ds ON ft.sku_key = ds.sku_key 
    WHERE ft.date_key BETWEEN 20080215 AND 20130107 
     AND ft.unit_cost_price BETWEEN 0 AND 1333 
     AND ft.unit_retail_price BETWEEN 0 AND 16500 
     AND ft.store_key IN ("1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15", "16") 
     AND ds.product_id IN (1312009,1312007,.. Huge List say 30000) 
     AND ds.category IN ("Male","Female","Unisex") 
     AND ft.day_of_week IN ("1","2","3","4","5","6","7") 
     AND ds.collection_name IN ("Base","SS12","AW12") 
     GROUP BY ds.product_id) AS t 
GROUP BY t.KeyValue 

Answer 1

1. Измерьте время запуска.

2. Измерьте время выполнения каждого запроса, входящего в состав оператора UNION.

3. Индекс каждой колонки, используемой в предложении WHERE.

4. Поместите наиболее избирательные столбцы сначала в предложение WHERE и проверьте влияние многоколоночных индексов.

5. Обрезать бесполезные тесты. (Конечно, ft.day_of_week можно удалить из предложений WHERE.)

6. Пересмотреть типы данных. Являются ли дни недели и хранят ключи действительно строки?

7. Пересмотреть решение о выборе пятилетних данных за раз.

8. Попробуйте переместить идентификационные номера продуктов во временную таблицу и присоединиться к ней.