Kevin,
if you answer my question below, I'll answer yours :) The answer for you is:
Bandwidth management ought to be done before the traffic hits the segment
that you want to manage, i.e. on the outgoing interface. That is, if you
want to throttle both ways, bandwidth management should be on both devices,
each handling the packets being sent.
For TCP/IP (but not for UDP/IP and other protocols), though, you can get
away by limiting the incoming and outgoing data stream on one side only.
This is because (the connection-oriented) TCP adapts to the data rate at
which it can send packets without packet loss. With UDP and others that are
"unreliable", packets will be sent anyway and get lost if the BCU decides to
discard them. We've found that in practice, if UDP and other protocols
aren't used too much, bandwidth shaping works quite well if applied on one
device only ... this includes the situation where the shaping device is
further up in the backhaul and does its job for more than one base station.
Now on to my question for you, which is related:
Karlnet can limit bandwidth on each interface. Does this bandwidth
limitation work only on the outgoing data stream of the interface where it
is applied? If yes, asymmetric bandwidth limiting should be possible like in
this example for a CPE:
Ethernet Interface 2400 kbps
Wireless Interface 256 kpbs
... this should give 2500/250 "DSL-like" asymmetric behavior, at least for
TCP. If UDP should be taken care of as well, the same policy needs to be
implemented at the base station or further up in the backbone.
However, if Karlnet bandwidth throttling works on both incoming and outgoing
packets on each interface, the net result would be the lower of the settings
- 256 kpbs - which is not really what was intended. So I hope your answer
will be "yes" to the first interpretation ...
Kind regards,
--
true global communications GmbH
Thomas Giger
In der Au 27, 61440 Oberursel, Germany
fon +49.6171.6381-0, fax +49.6171.6381-19
www.tgnet.de || www.megaspeed-internet.de
> -----Original Message-----
> From: Kevin Knuth [mailto:kknuth@karlnet.com]
> Sent: Tuesday, July 15, 2003 6:39 PM
> To: Karlnet Mailing List
> Subject: RE: [Karlnet] bandwidth management
>
>
> Time for me to learn something here:
>
> My question about Bandwidth Control Units is this: How do
> they limit the
> amount of bandwidth a customer gets? Does it actually stop
> the wireless
> device from trying to send data? I would think that is
> impossible. So if
> it does not actually control the CPE, aren't you creating more RF
> interference as units will continually try sending their
> packets to the AP?
>
> I have never really got a straight answer on this from
> anyone, so help will
> be appreciated.
>
> Kevin
>
> -----Original Message-----
> From: karlnet-bounces@WISPNotes.com
> [mailto:karlnet-bounces@WISPNotes.com]On Behalf Of Andy Henckel
> Sent: Tuesday, July 15, 2003 10:55 AM
> To: Karlnet Mailing List
> Subject: RE: [Karlnet] bandwidth management
>
>
> Once you get 250 customers on the same bridged wireless
> network, I'd say
> throttling at cpe is a must. One single unit presents a
> single point of
> failure, and the inability to isolate network problems to one
> site. We have
> been using YDI bandwidth manager for our wireless network.
> Without it,
> people can and will use all of the available bandwidth.
> Probably only a
> handful of our customers use the connection like that. It is
> essential to
> give everyone a chance to get bandwidth. We have recently
> added turbocell
> with one of many benefits being the ability to set speed on
> the CPE or on
> the BASE.
>
> -----Original Message-----
> From: LaRoy McCann [mailto:lmccann@roachconveyors.com]
> Sent: Tuesday, July 15, 2003 9:37 AM
> To: karlnet@WISPNotes.com
> Subject: [Karlnet] bandwidth management
>
>
> What are everyone's opinion on bandwidth control.
>
> Is it better to do it on each cpe or to have a single point
> at the head end
> to manage all bandwidth for users?
>
> What type of hardware/software is being used for single point
> bandwidth
> management?
>
> Thanks.
>
> _______________________________________________
> Karlnet mailing list
> Karlnet@WISPNotes.com
> http://lists.wispnotes.com/mailman/listinfo/karlnet
>
>
> _______________________________________________
> Karlnet mailing list
> Karlnet@WISPNotes.com
> http://lists.wispnotes.com/mailman/listinfo/karlnet
>
|