Re: [LSF][MM] rough agenda for memcg.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Mar 30, 2011 at 7:01 PM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
>
> Hi,
>
> In this LSF/MM, we have some memcg topics in the 1st day.
>
> From schedule,
>
> 1. Memory cgroup : Where next ? 1hour (Balbir Singh/Kamezawa)
> 2. Memcg Dirty Limit and writeback 30min(Greg Thelen)
> 3. Memcg LRU management 30min (Ying Han, Michal Hocko)
> 4. Page cgroup on a diet (Johannes Weiner)
>
> 2.5 hours. This seems long...or short ? ;)

I think it is a good starting plan.

> I'd like to sort out topics before going. Please fix if I don't catch enough.
>
> mentiont to 1. later...
>
> Main topics on 2. Memcg Dirty Limit and writeback ....is
>
>  a) How to implement per-memcg dirty inode finding method (list) and
>    how flusher threads handle memcg.

I have some very rough code implementing the ideas discussed in
http://thread.gmane.org/gmane.linux.kernel.mm/59707
Unfortunately, I do not yet have good patches, but maybe an RFC series
soon.  I can provide update on the direction I am thinking.

>  b) Hot to interact with IO-Less dirty page reclaim.
>    IIUC, if memcg doesn't handle this correctly, OOM happens.

The last posted memcg dirty writeback patches were based on -mm at the
time, which did not have IO-less balance_dirty_pages.  I have an
approach which I _think_ will be compatible with IO-less
balance_dirty_pages(), but I need to talk with some writeback guys to
confirm.  Seeing the Writeback talk Mon 9:30am should be very useful
for me.

>  Greg, do we need to have a shared session with I/O guys ?
>  If needed, current schedule is O.K. ?

We can contact any interested writeback guys to see if they want to
attend memcg-writeback discussion.  We might be able to defer this
detail until Mon morning.

> Main topics on 3. Memcg LRU management
>
>  a) Isolation/Gurantee for memcg.
>    Current memcg doesn't have enough isolation when globarl reclaim runs.
>    .....Because it's designed not to affect global reclaim.
>    But from user's point of view, it's nonsense and we should have some hints
>    for isolate set of memory or implement a guarantee.
>
>    One way to go is updating softlimit better. To do this, we should know what
>    is problem now. I'm sorry I can't prepare data on this until LSF/MM.
>    Another way is implementing a guarantee. But this will require some interaction
>    with page allocator and pgscan mechanism. This will be a big work.
>
>  b) single LRU and per memcg zone->lru_lock.
>    I hear zone->lru_lock contention caused by memcg is a problem on Google servers.
>    Okay, please show data. (I've never seen it.)
>    Then, we need to discuss Pros. and Cons. of current design and need to consinder
>    how to improve it. I think Google and Michal have their own implementation.
>
>    Current design of double-LRU is from the 1st inclusion of memcg to the kernel.
>    But I don't know that discussion was there. Balbir, could you explain the reason
>    of this design ? Then, we can go ahead, somewhere.
>
>
> Main topics on 4. Page cgroup on diet is...
>
>  a) page_cgroup is too big!, we need diet....
>     I think Johannes removes -> page pointer already. Ok, what's the next to
>     be removed ?
>
>  I guess the next candidate is ->lru which is related to 3-b).
>
> Main topics on 1.Memory control groups: where next? is..
>
> To be honest, I just do bug fixes in these days. And hot topics are on above..
> I don't have concrete topics. What I can think of from recent linux-mm emails are...
>
>  a) Kernel memory accounting.
>  b) Need some work with Cleancache ?
>  c) Should we provide a auto memory cgroup for file caches ?
>     (Then we can implement a file-cache-limit.)
>  d) Do we have a problem with current OOM-disable+notifier design ?
>  e) ROOT cgroup should have a limit/softlimit, again ?
>  f) vm_overcommit_memory should be supproted with memcg ?
>     (I remember there was a trial. But I think it should be done in other cgroup
>      as vmemory cgroup.)
> ...
>
> I think
>  a) discussing about this is too early. There is no patch.
>     I think we'll just waste time.
>
>  b) enable/disable cleancache per memcg or some share/limit ??
>     But we can discuss this kind of things after cleancache is in production use...
>
>  c) AFAIK, some other OSs have this kind of feature, a box for file-cache.
>     Because file-cache is a shared object between all cgroups, it's difficult
>     to handle. It may be better to have a auto cgroup for file caches and add knobs
>     for memcg.
>
>  d) I think it works well.
>
>  e) It seems Michal wants this for lazy users. Hmm, should we have a knob ?
>     It's helpful that some guy have a performance number on the latest kernel
>     with and without memcg (in limitless case).
>     IIUC, with THP enabled as 'always', the number of page fault dramatically reduced and
>     memcg's accounting cost gets down...
>
>  f) I think someone mention about this...
>
> Maybe c) and d) _can_ be a topic but seems not very important.
>
> So, for this slot, I'd like to discuss
>
>  I) Softlimit/Isolation (was 3-A) for 1hour
>     If we have extra time, kernel memory accounting or file-cache handling
>     will be good.
>
>  II) Dirty page handling. (for 30min)
>     Maybe we'll discuss about per-memcg inode queueing issue.
>
>  III) Discussing the current and future design of LRU.(for 30+min)
>
>  IV) Diet of page_cgroup (for 30-min)
>      Maybe this can be combined with III.
>
> Thanks,
> -Kame

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href


[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]