Comments on: Give Love to Your SSDs – Reduce innodb_io_capacity_max! https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/ Sat, 10 Feb 2024 00:42:15 +0000 hourly 1 https://wordpress.org/?v=6.5.2 By: weike.meng https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971566 Fri, 20 Dec 2019 07:06:43 +0000 https://www.percona.com/blog/?p=63830#comment-10971566 Hi Francisco Bordenave，thanks for your sharing.Look forward to your new posts!

]]> By: Francisco Bordenave https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971563 Thu, 19 Dec 2019 16:39:03 +0000 https://www.percona.com/blog/?p=63830#comment-10971563 Thanks Derek, indeed, a lot of code reading (credits on Yves here) was done to understand better some internals. In any case increasing io_cap_max with no reason may actually cause more problems than solutions, even performance problems.

]]> By: Francisco Bordenave https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971562 Thu, 19 Dec 2019 16:36:55 +0000 https://www.percona.com/blog/?p=63830#comment-10971562 weike.meng yeap, that’s exactly the reason that motivated this (and upcoming) post. There are more posts coming with extra details on flushing and how to calculate things properly. For now, I’d strongly recommend to decrease those values and start from defaults (which actually works pretty pretty well in most of cases)

]]> By: Derek Downey https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971561 Thu, 19 Dec 2019 16:24:00 +0000 https://www.percona.com/blog/?p=63830#comment-10971561 The MySQL I/O tuning docs mention keeping this (innodb_io_capacity) low depending on whether throughput is bottlenecked by flushing https://dev.mysql.com/doc/refman/5.7/en/optimizing-innodb-diskio.html .

This is a great post because the docs don’t make reference to innodb_io_capacity_max, or tying the increased (unnecessary) iops to reduced SSD lifecycle.

]]> By: Yves Trudeau https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971560 Thu, 19 Dec 2019 14:09:46 +0000 https://www.percona.com/blog/?p=63830#comment-10971560 Hi Federico, either it is because my PMM setup has non-standard scrape intervals or this is caused by the innodb_flushing_avg_loops (30s)

]]> By: weike.meng https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971557 Thu, 19 Dec 2019 10:53:48 +0000 https://www.percona.com/blog/?p=63830#comment-10971557 Before I read this post, I always set the innodb_io_capacity=10000 and innodb_io_capacity_max=20000 when the disk is ssd。 Now ,what can I do to mesure the value of innodb_io_capacity and the value of innodb_io_capacity_max？

]]> By: Federico Razzoli https://www.percona.com/blog/give-love-to-your-ssds-reduce-innodb_io_capacity_max/#comment-10971550 Thu, 19 Dec 2019 00:34:32 +0000 https://www.percona.com/blog/?p=63830#comment-10971550 Hi Yves and Francisco, great post.

The very first graph in the post is interesting. With your configuration flushed pages are constant, with down-spikes. They are:
* Irregular with innodb_io_capacity=300
* The height becomes regular with 200
* The interval between spikes also becomes regular with 300
Do you know the reason?

]]>