From: Nick Shah (nshah@connect.com.au)
Date: Tue Oct 08 2002 - 23:54:44 GMT-3
Chris,
You are right. I guess we should be lookign at the wording of the question.
That would be a giveaway. Plus our understanding of the underlying
technology.
For example. CBWFQ doesnt kick in until there is a congestion, when there is
a congestion you get the specified b/w (specified in the class/policy). It
also doesnt take care of serialization delay, jitter (what if the interface
is transmitting a large packet just around the time your voice packet comes
in, voice packet waits).
With RSVP, the required b/w is signalled at the beginning of the
conversation, if required b/w is not available tough luck, no call made.
This reservation is made end to end. And you may also need to specify
'req-qos guaranteed delay or req-qos controlled load' under dial peers.
With RTP priority, you use it when you are asked to strictly prioritise
voice and all else use CQ or WFQ. so you specify ip rtp priority <port>
<port> bw. Use it when you have to prioritise VOICE above all others.
And then there is frame relay fragmentation , if data & voice traffic takes
a single PVC path.
or DLCI prioritization along with PQ if using different PVC;s for
VOICE/DATA/whatever else.
Plus remember where the technologies can be implemented, on the edge or in
the core, for TCP or UDP. Like WRED can be used for TCP (TCP "adapts", UDP
doesnt), but for UDP its not really helpful.
We can virtually use any/every queuing technology depending on a scenario.
(WFQ even). It will all depend upon the wording of the question and what is
allowed/expected/disallowed etc.
Hope, I have not confused you more :) than you already were.
BTW, in your question, you cant use CBWFQ because the specified b/w in cbwfq
is 'minimum' amount of b/w usable during congestion, so rule it out of
equation.
You could use either RSVP or RTP.
rgds
Nick
----- Original Message -----
From: "Larson, Chris" <CLarson@usaid.gov>
To: <ccielab@groupstudy.com>
Sent: Wednesday, October 09, 2002 12:13 AM
Subject: RSVP
> I had a question in a lab scenario that called to make 40 k available to
> voice traffic. My first thought was ip rtp priority with a 40k max limit.
> However the solution calls for RSVP.
>
> Nowhere else is RSVP configured. Based on what I know about RSVP this
could
> not be correct. RSVP requires configuration all along the paths right? How
> would a person determine they are asking for RSVP? For that matter how
would
> you know if they are asking for CBWFQ? It could be used to accomplish the
> same thing if the acl were setup rightin the class map?
>
> I would think that if a solution called for RSVP the question would state
to
> reserve a certain amount of bandwidth, not make available. Thoughts? How
to
> determine what a given question wants in the way of QoS.
>
>
>
>
> Christopher Larson - CNE, MCP+I, CCNP + Security
> Open Systems Sciences Corp.
> USAID Information Resources Management (contractor)
> 202 712 4559
> clarson@usaid.gov
This archive was generated by hypermail 2.1.4 : Tue Nov 05 2002 - 08:35:43 GMT-3