Re: Maximize Number of Instances on RHEL 7.9
Date: Sun, 8 Oct 2023 09:13:18 -0700
Message-ID: <CAORjz=N1nTgA_qadyMvdd2nWCszFVJp3SGyTRR7cRzF9RdfRGw_at_mail.gmail.com>
On Sun, Oct 8, 2023, 8:55 AM <asterger_at_earthlink.net> wrote:
> Something I should have stated up front, all of these small instances
> are using AMM which is incompatible with Linux hugepages.
>
> Any other thing I have overlooked?
>
> -- Alan
>
> On 10/7/2023 9:44 PM, Jared Still wrote:
> > Something to keep in mind about recommendations: ask yourself why the
> > recommendation was made.
> >
> > As Mladen mentioned, hugepages are not swapped out.
> >
> > The load on the CPU is greatly reduced for memory operations.
> >
> > Standard memory pagesize on Linux is 4k.
> >
> > Run this cmd: getconf PAGESIZE
> > (I think that is correct)
> >
> > That reduces page lookups by 512x.
> >
> > With fewer addresses, the CPU TLB (Google it :) is more likely to
> > have it cached.
> >
> > I don't know why hugepages would not be recommended.
> >
> > Some OSs just do it automatically.
> >
> > On Sat, Oct 7, 2023, 10:32 AM <asterger_at_earthlink.net> wrote:
> >
> > Jared - Thanks for replying.
> >
> > From my research, hugepages was recommended for SGAs over 4G.
> > Most of
> > my SGAs are in the 1.2-3G range.
> >
> > Thanks,
> >
> > -- Alan
> >
> >
> > On 10/7/2023 12:40 PM, Jared Still wrote:
> > > One of the things I am curious about is why no hugepages?
> > >
> > > I'm not aware of hugepages being non-useful due to small SGA size.
> > >
> > > With so many instances I would think it rather a good feature to
> > employ.
> > >
> > >
> > > On Sat, Sep 2, 2023, 3:17 PM Alan Sterger
> > <asterger_at_earthlink.net> wrote:
> > >
> > > Hello,
> > >
> > > Looking for guidance in maximizing number of Oracle
> > instances on a
> > > RHEL
> > > 7.9 server for Dev/Val testing, not Production.
> > >
> > > We are running Oracle SE2 with threaded_execution and small
> SGAs
> > > therefore, not using server HugePages. Databases
> > > use_large_pages=false
> > > along with _ksmg_granule_size=4194304 (4M). Currently hosting
> 31
> > > instances, using memory_target with half of the instances set
> to
> > > 2560M,
> > > the other half at 1240M. Most are 19c, 6 at 12c.
> > >
> > > total used free shared buff/cache
> > > available
> > > Mem: 92G 10G 30G 35G
> > > 51G 45G
> > > Swap: 31G 0B 31G
> > >
> > > Assume I need to alter kernel parameters SHMMAX and SHMALL
> > from their
> > > defaults setting ~80% of the 92G for oracle databases.
> > >
> > > What can you add or suggest doing?
> > >
> > > Thanks,
> > >
> > > -- Alan
> > >
> > >
> > > --
> > > http://www.freelists.org/webpage/oracle-l
> > >
> > >
> >
>
>
-- http://www.freelists.org/webpage/oracle-lReceived on Sun Oct 08 2023 - 18:13:18 CEST