Re: [ATTEND] [LSF/MM TOPIC] Buffered writes throttling

Vivek Goyal [off-list ref] writes:

So, IIUC, the only thing little different here is that throttling is
implemented by flusher thread. But it is still per device per cgroup. I
think that is just a implementation detail whether we implement it
in block layer, or in writeback or somewhere else.  We can very well
implement it in block layer and provide per bdi/per_group congestion
flag in bdi so that flusher will stop pushing more IO if group on 
a bdi is congested (because IO is throttled).

I think first important thing is to figure out what is minimal set of
requirement (As jan said in another mail), which will solve wide
variety of cases. I am trying to list some of points. 


- Throttling for buffered writes
	- Do we want per device throttling limits or global throttling
	  limtis.

	- Exising direct write limtis are per device and implemented in
	  block layer.

	- I personally think that both kind of limits might make sense.
	  But a global limit for async write might make more sense at
	  least for the workloads like backup which can run on a throttled
  	  speed.

	- Absolute throttling IO will make most sense on top level device
	  in the IO stack.

	- For per device rate throttling, do we want a common limit for
	  direct write and buffered write or a separate limit just for
	  buffered writes.

Another aspect to this problem is 'dirty memory limiting'.  First a
quick refresher on memory.soft_limit_in_bytes...  In memcg the
soft_limit_in_bytes can be used as a way to overcommit a machine's
memory.  The idea is that the memory.limit_in_bytes (aka hard limit)
specified a absolute maximum amount of memory a memcg can use, while the
soft_limit_in_bytes indicates the working set of the container.  The
simplified equation is that if the sum(*/memory.soft_limit_in_bytes) <
MemTotal, then all containers should be guaranteed their working set.
Jobs are allowed to allocate more than soft_limit_in_bytes so long as
they fit within limit_in_bytes.  This attempts to provide a min and max
amount of memory for a cgroup.

The soft_limit_in_bytes is related to this discussion because it is
desirable if all container memory above soft_limit_in_bytes is
reclaimable (i.e. clean file cache).  Using previously posted memcg
dirty limiting and memcg writeback logic we have been able to set a
container's dirty_limit to its soft_limit.  While not perfect, this
approximates the goal of providing min guaranteed memory while allowing
for usage of best effort memory, so long as that best effort memory can
be quickly reclaimed to satisfy another container's min guarantee.

- Proportional IO for async writes
	- Will probably make most sense on bottom most devices in the IO
	  stack (If we are able to somehow retain the submitter's context).
	
	- Logically it will make sense to keep sync and async writes in
	  same group and try to provide fair share of disk between groups.
	  Technically CFQ can do that but in practice I think it will be
 	  problematic. Writes of one group will take precedence of reads
	  of another group. Currently any read is prioritized over 
	  buffered writes. So by splitting buffered writes in their own
	  cgroups, they can serverly impact the latency of reads in
	  another group. Not sure how many people really want to do
	  that in practice.

	- Do we really need proportional IO for async writes. CFQ had
	  tried implementing ioprio for async writes but it does not
	  work. Should we just care about groups of sync IO and let
	  all the async IO on device go in a single queue and lets
	  make suere it is not starved while sync IO is going on.


	- I thought that most of the people cared about not impacting
	  sync latencies badly while buffered writes are happening. Not
	  many complained that buffered writes of one application should
	  happen faster than other application. 

	- If we agree that not many people require service differentation
	  between buffered writes, then we probably don't have to do
	  anything in this space and we can keep things simple. I
	  personally prefer this option. Trying to provide proportional
	  IO for async writes will make things complicated and we might
	  not achieve much. 

	- CFQ already does a very good job of prioritizing sync over async
	  (at the cost of reduced throuhgput on fast devices). So what's
	  the use case of proportion IO for async writes.

Once we figure out what are the requirements, we can discuss the
implementation details.

Thanks
Vivek

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  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=mailto:"dont@kvack.org"> email@kvack.org </a>