Re: [PATCH 1/9] io-throttle documentation

[Gui Jianfeng <guijianfeng-BthXqXjhjHXQFUHtdCDX3A@public.gmane.org>]

KAMEZAWA Hiroyuki wrote:
> On Fri, 17 Apr 2009 15:34:53 +0800
> Gui Jianfeng <guijianfeng-BthXqXjhjHXQFUHtdCDX3A@public.gmane.org> wrote:
> 
>> KAMEZAWA Hiroyuki wrote:
>>> On Tue, 14 Apr 2009 22:21:12 +0200
>>> Andrea Righi <righi.andrea-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
>>>
>>>> +Example:
>>>> +* Create an association between an io-throttle group and a bio-cgroup group
>>>> +  with "bio" and "blockio" subsystems mounted in different mount points:
>>>> +  # mount -t cgroup -o bio bio-cgroup /mnt/bio-cgroup/
>>>> +  # cd /mnt/bio-cgroup/
>>>> +  # mkdir bio-grp
>>>> +  # cat bio-grp/bio.id
>>>> +  1
>>>> +  # mount -t cgroup -o blockio blockio /mnt/io-throttle
>>>> +  # cd /mnt/io-throttle
>>>> +  # mkdir foo
>>>> +  # echo 1 > foo/blockio.bio_id
>>> Why do we need multiple cgroups at once to track I/O ?
>>> Seems complicated to me.
>>   Hi Kamezawa-san,
>>
>>   The original thought to implement this function is for sharing a bio-cgroup
>>   with other subsystems, such as dm-ioband. If the bio-cgroup is already mounted,
>>   and used by dm-ioband or others, we just need to create a association between 
>>   io-throttle and bio-cgroup by echo a bio-cgroup id, just like what dm-ioband does.
>>
> 
> - Why we need multiple I/O controller ? 
> - Why bio-cgroup cannot be a _pure_ infrastructe as page_cgroup ?
> - Why we need extra mount ?
> 
> I have no answer but, IMHO, 
>  - only one I/O controller should be enabled at once.
>  - bio cgroup should be tightly coupled with I/O controller and should work as
>    infrastructure i.e. naming/tagging I/O should be automatically done by
>    I/O controller. not by the user's hand.

  It seems dm-ioband has to make use of bio-cgroup by the user's hand. Because dm-ioband
  is not cgroup based. :(

  Is that possible that another subsystem(not cgroup based, and not an IO Controller) also 
  would like to use bio-cgroup in the future? There's no such case at least now, so i don't 
  object to get rid of this part. :)

> 
> Thanks,
> -Kame
> 
> 
> 
> 
> 

-- 
Regards
Gui Jianfeng