Scalability Question

Scalability Question

[Nick Krause]

I am curious after reading some outdated kernel papers, how scalable
is the kernel of
late? I am curious mostly in memory and cpu subsystems as file systems
will change
based on user's choice.
Nicholas Krause

Re: Scalability Question

[Mike Galbraith]

On Fri, 2014-07-04 at 16:40 -0400, Nick Krause wrote: 
> I am curious after reading some outdated kernel papers, how scalable
> is the kernel of
> late? I am curious mostly in memory and cpu subsystems as file systems
> will change
> based on user's choice.

You can currently configure for up to 8192 CPUs.  I've not seen any
benchmark data whatsoever for huge boxen, have no idea where which
subsystem crumbles.  SGI asked for the increase to 8192, so presumably
do manage to squeeze acceptable performance out of size XXXXL boxen.

-Mike

Re: Scalability Question

[Nick Krause]

The most powerful super computer runs Ubuntu with over 3.2 million cores.
There fore I can state that Linux is very good at scaling as I have seem
the other side with embedded systems.
Cheers Nick

On Fri, Jul 4, 2014 at 11:24 PM, Mike Galbraith
<umgwanakikbuti@gmail.com> wrote:
> On Fri, 2014-07-04 at 16:40 -0400, Nick Krause wrote:
>> I am curious after reading some outdated kernel papers, how scalable
>> is the kernel of
>> late? I am curious mostly in memory and cpu subsystems as file systems
>> will change
>> based on user's choice.
>
> You can currently configure for up to 8192 CPUs.  I've not seen any
> benchmark data whatsoever for huge boxen, have no idea where which
> subsystem crumbles.  SGI asked for the increase to 8192, so presumably
> do manage to squeeze acceptable performance out of size XXXXL boxen.
>
> -Mike
>

Re: Scalability Question

[Richard Weinberger]

On Sat, Jul 5, 2014 at 5:33 AM, Nick Krause <xerofoify@gmail.com> wrote:
> The most powerful super computer runs Ubuntu with over 3.2 million cores.

These kind of computers don't run a single kernel.
See grid computing.

> There fore I can state that Linux is very good at scaling as I have seem
> the other side with embedded systems.
> Cheers Nick
>
> On Fri, Jul 4, 2014 at 11:24 PM, Mike Galbraith
> <umgwanakikbuti@gmail.com> wrote:
>> On Fri, 2014-07-04 at 16:40 -0400, Nick Krause wrote:
>>> I am curious after reading some outdated kernel papers, how scalable
>>> is the kernel of
>>> late? I am curious mostly in memory and cpu subsystems as file systems
>>> will change
>>> based on user's choice.
>>
>> You can currently configure for up to 8192 CPUs.  I've not seen any
>> benchmark data whatsoever for huge boxen, have no idea where which
>> subsystem crumbles.  SGI asked for the increase to 8192, so presumably
>> do manage to squeeze acceptable performance out of size XXXXL boxen.
>>
>> -Mike
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/



-- 
Thanks,
//richard

Re: Scalability Question

[Nick Krause]

That's true.
Cheers Nick

On Fri, Jul 4, 2014 at 11:44 PM, Richard Weinberger
<richard.weinberger@gmail.com> wrote:
> On Sat, Jul 5, 2014 at 5:33 AM, Nick Krause <xerofoify@gmail.com> wrote:
>> The most powerful super computer runs Ubuntu with over 3.2 million cores.
>
> These kind of computers don't run a single kernel.
> See grid computing.
>
>> There fore I can state that Linux is very good at scaling as I have seem
>> the other side with embedded systems.
>> Cheers Nick
>>
>> On Fri, Jul 4, 2014 at 11:24 PM, Mike Galbraith
>> <umgwanakikbuti@gmail.com> wrote:
>>> On Fri, 2014-07-04 at 16:40 -0400, Nick Krause wrote:
>>>> I am curious after reading some outdated kernel papers, how scalable
>>>> is the kernel of
>>>> late? I am curious mostly in memory and cpu subsystems as file systems
>>>> will change
>>>> based on user's choice.
>>>
>>> You can currently configure for up to 8192 CPUs.  I've not seen any
>>> benchmark data whatsoever for huge boxen, have no idea where which
>>> subsystem crumbles.  SGI asked for the increase to 8192, so presumably
>>> do manage to squeeze acceptable performance out of size XXXXL boxen.
>>>
>>> -Mike
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
>
>
>
> --
> Thanks,
> //richard