Re: precedence--protocol and reality

From: Howard C. Berkowitz (hcb@xxxxxxxxxxxx)
Date: Thu Apr 25 2002 - 10:17:29 GMT-3


   
There's an analogy I find very useful in discussing QoS: the
airlines. Airlines manage to work very nicely with essentially four
classes of service: crew, first, business, and economy. Where the
complexity comes in is the pricing, which may be different for
different customers.

It's like that for QoS, if you'll also permit me a medical analogy:
when you hear hoofbeats, the smart money is that they are horses, not
zebras. When you analyze the traffic mix of most organizations,
you'll find something like this with respect to QoS:

     network management
     delay-critical (VoIP, etc.)
     premium interactive (mission critical transaction processing)
     ordinary interactive (HTTP, Telnet, etc.)
     [possibly] premium bulk (perhaps mail over file transfer)
     bulk (FTP, SMTP, etc.)

That there is a three bit field for precedence reflects a particular
military requirement of the time rather than necessarily what is best
for QoS today. It's no more than a nice coincidence that the
practical number of levels of application precedence matches fairly
well.

Yes, there are a greater number of potential values of the DSCP, but
you'll actually find it is simpler when you consider that it is more
specific to QoS than IP precedence -- the bits mean something in QoS
terms rather than merely being descriptive. They split the
requirements for latency and loss. If you look at something like
VoIP, it's latency-critical but loss-tolerant. Synchronized database
update is of medium latency but loss-intolerant. Routing updates are
critical both for latency and loss.

At 6:36 AM -0400 4/25/02, Peter van Oene wrote:
>anything higher than normal should be fine.
>
>At 08:27 AM 4/25/2002 +0500, Ahmed Mamoor Amimi wrote:
>>So what should we use when we are told that prioritize traffice so that they
>>are escaped first.... 5 or 7.... what is our limit.
>>
>>-Mamoor
>>----- Original Message -----
>>From: Peter van Oene <pvo@usermail.com>
>>To: Ahmed Mamoor Amimi <mamoor@ieee.org>; <ccielab@groupstudy.com>
>>Sent: Wednesday, April 24, 2002 11:38 PM
>>Subject: Re: precedence
>>
>>
>>> control traffic generally flows with 111. Hence, if you set your data
>>> traffic to 111, your routers with no longer be able to prioritize control
>>> packets which can have rather negative effects when the dropping
>>> begins. Many times 111 and 110 are also handled with equal priority which
>>> may be the reason why cisco advises against both.
>>>
>>> Pete
>>>
>>>
>>> At 08:07 PM 4/24/2002 +0500, Ahmed Mamoor Amimi wrote:
>>> >Can some expert tell me why it is not recommended precedence value of 6
>>and 7.
>>> >I have seen at CCO that it will produce potential problem with the core
>>data
>>> >traffic like routing and other signalling.... if this is so then why
>>these
>> > >value are given to be configured.

--
"What Problem are you trying to solve?"
***send Cisco questions to the list, so all can benefit -- not
directly to me***
*******************************************************************************
*
Howard C. Berkowitz      hcb@gettcomm.com
Chief Technology Officer, GettLab/Gett Communications http://www.gettlabs.com
Technical Director, CertificationZone.com http://www.certificationzone.com
"retired" Certified Cisco Systems Instructor (CID) #93005


This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:58:19 GMT-3