Re: Anonymous huge pages for Linux
Date: Mon, 9 Jan 2023 18:25:16 -0500
Message-ID: <CAKVQkW6TQmujv-x_0V2zrr25am5b63MuKypEhS=UhbZBbOxWGA_at_mail.gmail.com>
Jon,
My apologies for the sharp NO. It was an echo from when a bunch of us
talked about this over a night of DBA debauchery. Mark, thanks for the
better answer.
My 2-cents. Consider configuring your Oracle DB server build policy to set
huge pages to 40% of physical memory. This 40% is for the *combined* SGAs
of all the databases on the database server. The other 60% is for the
*combined
I hope this is also a better answer.
Best of luck on your Oracle adventures.
Todd
On Mon, Jan 9, 2023 at 5:26 PM Mark J. Bobak <mark_at_bobak.net> wrote:
> To the best of my knowledge, anonymous husepages are still a non-starter
*PGA Aggregate Limits (*not targets*) of all the databases on that server
and OS. 40% is just a suggestion, 35% may be better. Do the math for what
works best for your environment.
> when it comes to Oracle.
>
> On Mon, Jan 9, 2023 at 3:44 PM Jon Crisler <joncrisler_at_gmail.com> wrote:
>
>> Has anybody had success using anonymous huge pages in later versions of
>> Red Hat or OEL ? The accepted practice has been to disable anonymous huge
>> pages and use fixed huge pages, due to performance issues and bugs.
>> However we are working on a project that may require the rapid provision
>> of small test databases, which will make maintaining manual huge pages a
>> headache. If anonymous huge pages work well now, that would be my
>> preferred configuration.
>>
>
-- http://www.freelists.org/webpage/oracle-lReceived on Tue Jan 10 2023 - 00:25:16 CET