基本插入/更新性能问题

Question

Running MySQL Database and all tables are InnoDB

Every evening we have a list of the top offending slow queries emailed to us via a batch script.

In this list are two queries an insert and an update.

The insert script logs page views for a user. To make this more efficient we will only write an insert once for a user in a session. So we store the pages viewed in a session and do an in_array on page load. If the page is not in the array than we write an insert.

This table has 700,000+ records with the only index being the primary key (id).

The insert into this table looks like this:

insert into page_views (user_id,page_name,created)
values ('4','Test Page','2013-08-13 10:44:21')

This query shows up on our slow query log as taking a while. My question is how can this be made more efficient. I feel as if this is part of the reason our database is running slower.

---

We also have an update query that runs on an InnoDB table with about 14,000+ records. The update happens when a user views a specific record for the first time and we set a field titled viewed='YES'.

update test_table set viewed='YES' where id=55

This update is showing up on the slow query log as well. This table "test_table" has about (5) different indexes on it.

---

#

Output from slow query log

Count : 165 (0.35%) Time : 2252.210041 s total, 13.649758 s avg, 2.026723 s to 313.311842 s max (8.96%) 95% of Time : 1181.480258 s total, 7.573591 s avg, 2.026723 s to 49.99961 s max Lock Time (s) : 12.722 ms total, 77 avg, 38 to 233 max (0.04%) 95% of Lock : 11.681 ms total, 75 avg, 38 to 84 max Rows sent : 0 avg, 0 to 0 max (0.00%) Rows examined : 0 avg, 0 to 0 max (0.00%) Database : Users : XXXXX@localhost : 100.00% (165) of query, 99.76% (46921) of all users

Query abstract: SET timestamp=N; INSERT INTO page_views (user_id, page_name, created) VALUES ('S', 'S', 'S')1;

The table structure for page_views

id (int) primary key page_name (varchar(255)) created datetime

What are my options for speeding up simple insert and updates on very large tables?

Answer 1

Based on the huge number of rows in your table and rows should be added/modified regularly in those tables. I'd suggest to execute following query first and see the execution time for the queries you have mentioned:

analyze table page_views;

analyze table test_table;

Above commands will not take much time, should be executed with 5/10 secs.

If you can afford little longer maintenance time(depends on the size of table):

optimize table page_views;

optimize table test_table;

If you see performance gain, I also would like to inform that you can configure cron job which executes analyze daily and optimize weekly/monthly.