Question

The table arg_rec contains 800K rows on my test machine, normally this table will hold over 15M rows. I want to run the following query :

SELECT STE_ID, PNT_NO, YR, MN, AVG(AVR_WS) AS AVR_WS, SUM(AVR_PW) FROM arg_rec GROUP BY STE_ID, PNT_NO, YR, MN;

This query gives daily average wind speed and total power from wind turbine data. On my test machine this query times out after 10 minutes of execution with a combined index on STE_ID, PNT_NO, YR, MN, which is only a subset of the primary key columns. Without the index the query completes after several minutes.

I am running with pretty much a stock MySQL installation, in addition to tweaking the server I would also like to know more about other ways to handling this problem, such as :

1. Is it possible to create a view based on this query and cache the results ?
2. Are there more advanced indexing features to encapsulate the fact that YR, MN, DY, HR, MI, SC correspond to year, month, day etc field of the record timestamp ?
3. Am I better off just duplicating the data using the business layer of my application ?

Answer 1

For best performance in GROUP BY queries you must add covering index as:

ALTER TABLE arg_rec ADD KEY ix1(STE_ID, PNT_NO, YR, MN,AVR_WS, AVR_PW );

For covering index you add:

1. columns used in where clauses first, then
2. columns used in group by, then
3. columns used in order by, and then
4. columns used in select.

Visit for details: Group By Optmization in MySQL

1. You can cache queries in MySQL by enabling query cache loot at Query Cache Configuration

2. You can store YR, MN, DY, HR, MI, SC in a single column having data type as TIMESTAMP which will increase performance of indexing and group by operation.