Re: Experiences & recommendations related to HUGEPAGES in RHEL 7

From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Sat, 15 Jul 2017 23:37:56 -0400
Message-ID: <20170715233756.454efc9d_at_umajor.home.com>


On Sat, 15 Jul 2017 12:27:04 -0400
Dba DBA <oracledbaquestions_at_gmail.com> wrote:

> We use huge pages here as a standard. How do people test huge pages to
> isolate variables so you can measure the improvement from huge pages?

Basically, huge pages should result in less time being spent in kernel mode. The only real consequences of huge pages are smaller page tables and SGA not being handled by swapper and page stealing daemon. In case of virtual machines, paging is even more expensive, so smaller page tables are even more significant. You are allocating the same amount of memory either using 4K pages or using 2M pages. The only real difference is the amount of overhead work that system has to perform on this memory. And in case of huge pages, that overhead is few percents lower than in the case of the 4K pages. That's about it. It's nothing spectacular, it's not a "turbo" button. You will not turn on the afterburner by using huge pages. Regards

-- 
Mladen Gogala
Oracle DBA
Tel: (347) 321-1217
--
http://www.freelists.org/webpage/oracle-l
Received on Sun Jul 16 2017 - 05:37:56 CEST

Original text of this message