General list for discussing Bufferbloat
 help / color / mirror / Atom feed
* [Bloat] wifi AP switching time
@ 2013-01-01 17:49 Michael Richardson
  2013-01-03  9:18 ` Pedro Tumusok
  0 siblings, 1 reply; 5+ messages in thread
From: Michael Richardson @ 2013-01-01 17:49 UTC (permalink / raw)
  To: bloat


A comment was made a month ago or so about how long it takes wifi APs
to switch from transmitting (unicast) to one station to another.  That
there was quite a large latency here, and that this was one reason that
the AP designers wanted large buffers to accmulate, so that the
switching time could be amortized over a larger number of packets.

I'm looking for a definitive reference to this problem.
Will this be in the 802.11 specification, or is this an inherent problem
in the chipsets, not the physics?

-- 
]       He who is tired of Weird Al is tired of life!           |  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
   Kyoto Plus: watch the video <http://www.youtube.com/watch?v=kzx1ycLXQSE>
	               then sign the petition. 



^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bloat] wifi AP switching time
  2013-01-01 17:49 [Bloat] wifi AP switching time Michael Richardson
@ 2013-01-03  9:18 ` Pedro Tumusok
  0 siblings, 0 replies; 5+ messages in thread
From: Pedro Tumusok @ 2013-01-03  9:18 UTC (permalink / raw)
  To: Michael Richardson; +Cc: bloat

[-- Attachment #1: Type: text/plain, Size: 878 bytes --]

On Jan 1, 2013 6:50 PM, "Michael Richardson" <mcr@sandelman.ca> wrote:
>
>
> A comment was made a month ago or so about how long it takes wifi APs
> to switch from transmitting (unicast) to one station to another.  That
> there was quite a large latency here, and that this was one reason that
> the AP designers wanted large buffers to accmulate, so that the
> switching time could be amortized over a larger number of packets.
>
> I'm looking for a definitive reference to this problem.
> Will this be in the 802.11 specification, or is this an inherent problem
> in the chipsets, not the physics?
>

I might be way of here, but to me this sounds like a-mpdu which is used to
aggregate frames to get higher throughput. Thats in the specification, its
in 802.11ac and I believe that it go introduced with 802.11n.

It is there to mitigate the overhead of aquiring the channel.

[-- Attachment #2: Type: text/html, Size: 1065 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bloat] wifi AP switching time
  2013-01-07  1:34   ` Michael Richardson
@ 2013-01-07  3:43     ` Pedro Tumusok
  0 siblings, 0 replies; 5+ messages in thread
From: Pedro Tumusok @ 2013-01-07  3:43 UTC (permalink / raw)
  To: Michael Richardson; +Cc: davecb, bloat

[-- Attachment #1: Type: text/plain, Size: 1128 bytes --]

On 7 Jan 2013 09:35, "Michael Richardson" <mcr@sandelman.ca> wrote:
>
>
> I asked
>     > A comment was made a month ago or so about how long it takes wifi
APs
>     > to switch from transmitting (unicast) to one station to another.
 That
>     > there was quite a large latency here, and that this was one reason
that
>     > the AP designers wanted large buffers to accmulate, so that the
>     > switching time could be amortized over a larger number of packets.
>
> Pedro Tumusok <pedro.tumusok@gmail.com> replied
>     > I might be way of here, but to me this sounds like a-mpdu which is
>     > used to aggregate frames to get higher throughput. Thats in the
>     > specification, its
>     > in 802.11ac and I believe that it go introduced with 802.11n.
>     >
>     > It is there to mitigate the overhead of aquiring the channel.
>
> Does this mean that 802.11a,b do not suffer from this problem?
>

They do not have that feature, but I assume they would still be bloated in
other places for other reasons.

The a-mpdu got tweaked a bit in ac, to always use it. Even for single
frames. Not only for aggregate as in n.

[-- Attachment #2: Type: text/html, Size: 1474 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bloat] wifi AP switching time
  2013-01-04 15:08 ` David Collier-Brown
@ 2013-01-07  1:34   ` Michael Richardson
  2013-01-07  3:43     ` Pedro Tumusok
  0 siblings, 1 reply; 5+ messages in thread
From: Michael Richardson @ 2013-01-07  1:34 UTC (permalink / raw)
  To: davecb; +Cc: bloat


I asked
    > A comment was made a month ago or so about how long it takes wifi APs
    > to switch from transmitting (unicast) to one station to another.  That
    > there was quite a large latency here, and that this was one reason that
    > the AP designers wanted large buffers to accmulate, so that the
    > switching time could be amortized over a larger number of packets.

Pedro Tumusok <pedro.tumusok@gmail.com> replied
    > I might be way of here, but to me this sounds like a-mpdu which is
    > used to aggregate frames to get higher throughput. Thats in the
    > specification, its 
    > in 802.11ac and I believe that it go introduced with 802.11n.
    > 
    > It is there to mitigate the overhead of aquiring the channel.

Does this mean that 802.11a,b do not suffer from this problem?

-- 
]               Never tell me the odds!                 | ipv6 mesh networks [ 
]   Michael Richardson, Sandelman Software Works        | network architect  [ 
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [ 
	

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bloat] wifi AP switching time
       [not found] <mailman.5.1357243201.32365.bloat@lists.bufferbloat.net>
@ 2013-01-04 15:08 ` David Collier-Brown
  2013-01-07  1:34   ` Michael Richardson
  0 siblings, 1 reply; 5+ messages in thread
From: David Collier-Brown @ 2013-01-04 15:08 UTC (permalink / raw)
  To: bloat

On Jan 1, 2013 6:50 PM, "Michael Richardson" <mcr@sandelman.ca> wrote:
>> A comment was made a month ago or so about how long it takes wifi APs
>> to switch from transmitting (unicast) to one station to another.  That
>> there was quite a large latency here, and that this was one reason that
>> the AP designers wanted large buffers to accmulate, so that the
>> switching time could be amortized over a larger number of packets.

Pedro Tumusok <pedro.tumusok@gmail.com> replied
> I might be way of here, but to me this sounds like a-mpdu which is used to
> aggregate frames to get higher throughput. Thats in the specification, its
> in 802.11ac and I believe that it go introduced with 802.11n.
> 
> It is there to mitigate the overhead of aquiring the channel.

And that in turn sounds like a place where codel may be needed to ensure
the buffer drains down to nothing in the general case. Alternatively, a
simpler approach might be to buffer only when actively trying a acquire
a channel.

I wonder if there are other only-for-one-reason buffers lying around in
the communication path (;-))

--dave
-- 
David Collier-Brown,         | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
davecb@spamcop.net           |                      -- Mark Twain
(416) 223-8968

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2013-01-07  3:43 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-01-01 17:49 [Bloat] wifi AP switching time Michael Richardson
2013-01-03  9:18 ` Pedro Tumusok
     [not found] <mailman.5.1357243201.32365.bloat@lists.bufferbloat.net>
2013-01-04 15:08 ` David Collier-Brown
2013-01-07  1:34   ` Michael Richardson
2013-01-07  3:43     ` Pedro Tumusok

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox