Server product kernel requirements

Prarit Bhargava prarit at redhat.com
Thu Oct 31 12:36:59 UTC 2013



On 10/30/2013 07:32 PM, Josh Boyer wrote:
> On Wed, Oct 30, 2013 at 7:25 PM, Prarit Bhargava <prarit at redhat.com> wrote:
>> On 10/30/2013 02:10 PM, Simo Sorce wrote:
>>> On Wed, 2013-10-30 at 10:51 -0700, David Strauss wrote:
>>>> On Wed, Oct 30, 2013 at 10:09 AM, Josh Boyer <jwboyer at fedoraproject.org> wrote:
>>>>> Massive 4096 multi-cored CPU machines with terabytes of DRAM and
>>>>> petabytes of storage, or more commodity style hardware used in
>>>>> heterogeneous environments, etc.
>>>>
>>>> The latter. We'd want a separate HPC group for 512+ core machines.
>>>
>>> Or simply, sites so big can care for their own kernel builds most
>>> probably, or seek for commercial support.
>>
>> Why limit it so low?  If we're thinking about going big, well, GO BIG.
>>
>> Users of Fedora want to support these systems out-of-the-box so they can get an
>> idea if their systems work.  Stopping at 512 just seems too low these days.
>>
>> We're talking about saving a very small amount of memory by not going to 4096 ..
> 
> Remind me how much again?  IIRC, it was around 2MB additional runtime
> overhead to set MAX_CPUS to that, right?  That's very small on
> servers, not so small on cloud.

Right, I think that was about it... it may be a little less than that.  I
wonder, however, how many people are actually using a bleeding-edge fedora
kernel for memory-critical cloud purposes?  I have a feeling that it's in the
same order of magnitude of people booting fedora on systems with greater than
512 cpus.

P.

> 
> josh


More information about the kernel mailing list