What vexes VoIP users? - Bufferbloat
jg at freedesktop.org
Thu Mar 3 16:17:28 CST 2011
On 03/01/2011 04:32 AM, William Pitcock wrote:
> That is the same market Vonage is now targeting in the US, basically.
> National calling in the US is basically bundled with most calling plans
> now. I'm not convinced that many people use Vonage in the US - my
> experience with it was that it was not as reliable as the VOIP
> products offered through the various broadband providers I have had.
Due to bufferbloat in the broadband edge, the broadband carriers have a
fundamental advantage in providing VOIP, since they do not do so over
the data service the user has but does not have access to for any
classification; it is provisioned entirely separately on different channels.
As you can see in the ICSI Netalyzr data you can find in my blog at
whenever a home connection is saturated for any reason, customers can
easily experience *seconds* of latency. (Telephony standards for max
latency + jitter are in the 150MS range). Even web browsing induces
transient jitter of order hundred(s) of milliseconds, from some
experiments I've done, which is a problem for VOIP, much less the bulk
data transfers which kill you for long periods.
Now that I have mitigated the bufferbloat disaster in my home cable
service via bandwidth shaping, Skype works sooo much better for me.
This is what devices such as Ooma are doing. Unfortunately, it means
you have to defeat features such as Comcast's PowerBoost.
Note I do not believe bufferbloat was intended by any broadband carrier
to give them such an advantage. Right now, they take it in the ear on
service calls. And as far as I've been able to tell, just about
everyone has been making the same generic mistake. I'm sure the
conspiracy theorists will love to make such claims, however.
If you don't know what bufferbloat is, you can try the talk I gave
recently in Bell Labs, available at:
http://mirrors.bufferbloat.net/Talks/BellLabs01192011/ or wade through
my blog at: http://gettys.wordpress.com/ or come to the transport area
meeting at the Prague IETF where I will be giving a somewhat abbreviated
version of the talk.
More information about the NANOG