Redis是否应该缓存数据库条目？

Question

My server is Debian 9.0. It has SSD HDD with 32 of RAM. and i7 processor with 4 (8) cores. I have 4 websites. 2 of them are very small and two large. The large databases (mysql) are about 120 + 60 MB in size. My webserver is nginx, also using varnish.

Well, i want to learn about linux so i started reading articles about tuning nginx, mysql, php-fpm etc. I was tuning mysql, installed mysqltuner. It advised me to set "query_cache_size" and "query_cache_type" to 0 which means disabling the cache..

I did. However my sites become really slower. My 680 Kb web page was loading in 700 ms but after disabling the query_cache, the speed decreased to almost 1,5 seconds. It was interesting. I set the query_cache_size back to 128M. So the speed was restored. I kept reading more about setting query_cache and read the article of Zaitsev (in percona.com) where he states that using a large cache is problematic and he says at the end of his article: " In many cases you may want to disable query cache at all and use external cache such as memcached which does not have this problem."

https://www.percona.com/blog/2007/03/23/beware-large-query_cache-sizes/

So, i installed REDIS (which they say is better than memcached and does things better). I installed it, i also installed the wordpress plugin for redis. i monitor redis and it works very fine. However two things happen:

1) Redis didnt improve my site speed. Its the same. No difference. Perhaps its because my web pages are very small (700 kb at most) and my databases are small also.

2) When i disable mysql query cache, my WEB SITE gets very slow again.

So isnt redis supposed to cache all DB entries and use them much faster than mysql? How come my webpage speed decreases when i disable query cache of mysql?

Answer 1

I agree that 700ms is too long. Properly designed, you should be able to execute a hundred MySQL queries in that time; sounds like you aren't?

I agree that a large Query cache is usually unwise:

• It takes away RAM from other caching.
• Every modification to a table cause purging of all entries in the QC for that table. If a table is changing often, you suffer overhead for that purging, while getting little or no benefit from caching.

You have a modest sized database -- under 1GB for everything? With 32 GB of RAM, you should be able to cache all the data and indexes all the time. So, slowdowns will be in poor indexing and/or poor query formulation.

I hope you are using InnoDB, not MyISAM. With InnoDB, you need to set innodb_buffer_pool_size to no more than 70% of available RAM. You have several products running in the 32G?? So, I won't say 70% of 32. (Note: Swapping is terrible for performance.) Since you have so little data, perhaps 2G would be a good setting?

Let's see the queries in your "too slow" web page. Also, let's see SHOW CREATE TABLE for the tables. Do you know about "composite" indexes? Do you know not to hide indexed columns in function calls?

Perhaps my cookbook can help you from here. If you are using WP, it includes a section on how to speed up wp_postmeta . If you need more help, start another Question with the slowest query, its EXPLAIN , and its schema.