[ale] Diagnostic help needed

Matt Rutherford matthew.g.rutherford at gmail.com
Sat Jan 14 22:17:57 EST 2012


Trying not to run on to much but here is an overview of how DOCSIS does
dynamic bandwidth. I'll leave out the more detailed guts of it. An in-depth
of this subject is way too OT here.

DOCSIS has a feature that allows for burst speed provisioning within a
specified range when there is bandwidth availability on the server. This
can be applied to a set amount of a specific transaction, ie: 5MB, 10MB,
20MB, etc. This provides the nice bump in initial download/buffer speed on
large videos (see Youtube, Netflix, etc) and also is quite noticeable on
speed test sites since they are file-transfer speed based. Also looks great
for marketing and advertising in comparison to the speeds offered on DSL.
This is all handled by the CMTS.

Typically the burst speed will be set to the around the same level as your
provisioned speed, though this is not always true. IE: a 22mbps connection
will burst speeds up to 44mbps (down) for a pre-determined amount of
upwards or downwards transfer. Since cable speeds are non-synchronous, you
may have a 5mbps upstream with bursts up to 10mbps. This cannot be relied
on however, hence the careful use of 'up to (X)mbps' in advertising.

Peak usage hours for residential areas for cable internet bandwidth run
5pm-midnight as Kirsa said, so I'd add a +1 to the congestion theory. There
are other causes as well, but tech support should be checking for those
symptoms anytime you call in (SNR, bad RX/TX, and high levels of error
correction to packets from the modem).

During heavy usage if there is saturation of the bandwidth availability to
below a specified threshold, protocols kick in to limit users to speeds
lower than their 'provisioned' default. This is done dynamically since each
CMTS has multiple RF cards separated into upstream and downstream. You may
have a non-saturated upstream but a saturated or congested downstream as
far as bandwidth goes - we rarely saw upstream problems except for military
barracks, college towns, and other edge cases. There are other layers of
bandwidth management and bottlenecks that impact speeds which can exist at
the node or network level, but outside of RF issues the most typical and
noticeable change in speeds comes from this type of congestion.

-Matt R


On Sat, Jan 14, 2012 at 8:48 PM, Jim Kinney <jim.kinney at gmail.com> wrote:

> All ISPs know the speedtest sites. Comcast is known to privide priority
> routing.
> On Jan 14, 2012 7:53 PM, "Drifter" <drifter at oppositelock.org> wrote:
>
>> On Saturday, January 14, 2012 05:36:53 pm Kirsa wrote:
>> > I would not be surprised if you experienced a similar speed issue next
>> > Friday though... On almost any residential cable node Friday 5-12pm is
>> > peak time and has the most utilization of the entire week. So that you
>> > were down to 4mbit/s on Friday evening and 2mbit/s by Friday night
>> > sounds suspiciously like congestion to me, despite the faulty
>> > equipment.
>> [Rest trimmed for brevity.]
>>
>> I had the same speed issues at 9 am Saturday morning. Snail-slow download
>> speeds continued all day Saturday until I replaced the defective hub.
>> Then: BINGO! full speed returned.  In fact, a speed about 19:45 Saturday
>> (Speakeasy's test) returned the rather unbelievable download speed of
>> 40.59 mbs! The upload speed was only 4.16. Obviously some sort of burst
>> speed. Does Comcast have some way to know when I access a speed test site
>> and goose the speed for a few seconds?
>>
>> Sean
>> _______________________________________________
>> Ale mailing list
>> Ale at ale.org
>> http://mail.ale.org/mailman/listinfo/ale
>> See JOBS, ANNOUNCE and SCHOOLS lists at
>> http://mail.ale.org/mailman/listinfo
>>
>
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://mail.ale.org/mailman/listinfo/ale
> See JOBS, ANNOUNCE and SCHOOLS lists at
> http://mail.ale.org/mailman/listinfo
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.ale.org/pipermail/ale/attachments/20120114/7181933b/attachment.html 


More information about the Ale mailing list