Message-ID: <20100831184030.GB9305@gvim.org>
Date:	Tue, 31 Aug 2010 11:40:30 -0700
From:	mark gross <markgross@...gnar.org>
To:	Kevin Hilman <khilman@...prootsystems.com>
Cc:	markgross@...gnar.org, Saravana Kannan <skannan@...eaurora.org>,
	linux-kernel@...r.kernel.org, linux-arm-msm@...r.kernel.org,
	"Rafael J. Wysocki" <rjw@...k.pl>,
	James Bottomley <james.bottomley@...e.de>,
	Frederic Weisbecker <fweisbec@...il.com>,
	Jonathan Corbet <corbet@....net>
Subject: Re: [PATCH] pm_qos: Add system bus performance parameter

On Mon, Aug 30, 2010 at 11:56:54AM -0700, Kevin Hilman wrote:
> mark gross <markgross@...gnar.org> writes:
> 
> > On Fri, Aug 27, 2010 at 07:55:37PM -0700, Saravana Kannan wrote:
> >> mark gross wrote:
> >> >On Fri, Aug 27, 2010 at 01:10:55AM -0700, skannan@...eaurora.org wrote:
> >> >>Ignoring other details for now, the biggest problem with throughput/KBps
> >> >>units is that PM QoS can't handle it well in its current state. For KBps
> >> >>the requests should be added together before it's "enforced". Just picking
> >> >>the maximum won't work optimally.
> >> >
> >> >well then current pm_qos code for network throughput takes the max.
> >> 
> >> I don't know how the network throughput is enforced, but if the unit
> >> is KBps and it's just doing a Max, then I think it's broken. If two
> >> clients request 50 KBps and your network can go till 200 KBps, you
> >> would still be requesting 50 KBps when you could have requested 100
> >> KBps.
> >> 
> >> Any specific reason PM QoS doesn't support a "summation" "comparitor"?
> >
> > PM_QoS could do a summation, but keep in mind it pm_qos not qos.  pm_qos
> > is a best effort thing to constrain power management throttling, not
> > provide a true quality of service or deadline scheduling support.
> 
> For me (and I think Saravana too), this is still all about power, but
> it's closely tied to QoS.  
> 
> For things like busses, which are inherently shared, PM is tightly
> coupled with "true" QoS, so I'm not sure I fully follow the distinction
> being made between PM QoS and QoS.  Seems like the tradeoff is always
> between power and performance.
> 
> > If you stick to the full up quality of service mentality you quickly get
> > into discussions just like those around memory over commit.  Its really
> > hard to know when best effort or hard QoS is appropriate.
> >
> > If you are trying to use pm_qos as a true qos interface then, its
> > definitely not up to the task.
> >
> > example:  you have one 100Mb NIC in your box.  With PM QoS you could
> > have 4 user mode applications requesting 100Mbs PM_Q0S.  In this case
> > the right thing to do is to constrain the NIC PM to keep it full on and
> > the PHY going as fast as it can.  But you'll never get 400Mbs out of the
> > thing.
> >
> > So far only max and min really have made sense for pm_qos but, if a case
> > pops up where summation makes more sense for aggregating the pm_qos
> > requests then I'm open to it.
> 
> Using your example above, what if the 4 apps all request 10Mb/s?
> 
> What is best effort?  Leave the NIC in 10Mb/s mode, or bump up the power
> state to 100Mb/s mode?

Now I get it!  For throughput we need to do a sum.  Ok, we need sum
comparator/performance aggregaters too!

Do we also need to figure out the max throughput and warn if the pm_qos
requests are going over?  I suppose the network stack could register
each device with a max bus bandwidth and pm_qos could warn on exceeding
the hardware throughput.

> This decision is both QoS and PM related.  Without summation, the 'max'
> request is still 10Mb/s so you would keep the lower power state.  But
> you also know that none of the clients will get their requested rate.
> 
> There's some gray area here since there is a choice.  Was the point
> of the request to keep the NIC at the *power-state* needed for 10Mb/s (a
> PM request) or was the request saying the app wanted at least 10Mb/s (a
> QoS request.)  

I need to think on this a bit.  You are correct, and it looks like we
could use both types of interfaces.

> 
> My understanding is that PM QoS is intended to limit power-state
> throttling.  IOW, in the absence of PM QoS requests, the PM core code is
> free to throttle the power of the devices/subsystems/busses etc.  If
> requests are present, it is no longer free to throttle blindly.
> 
> The question here seems to be whether or not the PM core code should
> also be free to increase the power state to meet a combination of PM QoS
> requests.  To me this is still PM related.  Just like in race-to-idle
> for the CPU, it might be better for overall power to go to the highter
> state for a burst and then be able to fully throttle again.
> 
> Kevin

thanks for the example!  it really helped me to understand the issue
better.

--mark

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists