From: SIMON HART (simon.hart@btinternet.com)
Date: Wed Feb 23 2005 - 07:16:47 GMT-3
To Amran and everyone concerned,
The company I work for is not BT - I repeat not BT. The email address I use for this bulletin board bears no reflection on the company I work for - (nor is it Thus!!)
Simon
"Ahmed, Amran" <Amran.Ahmed@let-it-be-thus.com> wrote:
Simon
Not much point in keeping your company anonymous when your email identifies
you.. ;-)
Amran
-----Original Message-----
From: simon hart [mailto:simon.hart@btinternet.com]
Sent: 22 February 2005 20:48
To: Roy Dempsey; swm@emanon.com
Cc: Cisco certification
Subject: RE: Setting IP precedence in the real world?
Roy,
You are correct in your assumption that if an ISP does nothing with the
IPprec they receive from their clients and they have WFQ and WRED enabled on
their interfaces then you could potentially derive some benefit. However
the major assumption is that they do nothing with the IPprec.
I cannot speak for every ISP out there, however the company I work for (they
shall remain nameless for the time being) will re-mark all traffic on the
ingress Access router. However things do not stop there. Our core network
does not switch 'raw' IP packets, it uses MPLS for connectivity across our
ISP core. If we are offering 'vanilla' ISP services then there would be no
IPprec, or Diffserv to MPLS exp bit mapping. Therefore the core routers
would not have an inkling on how the IP Packets are marked.
Therefore if you were to use the ISP services of the company I work for,
then you would derive no benefit from marking your IP packets. It becomes
somewhat different if you were to subscribe to an MP-iBGP MPLS VPN - but I
guess that is another subject
Simon
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of Roy
Dempsey
Sent: 22 February 2005 18:02
To: swm@emanon.com
Cc: Cisco certification
Subject: Re: Setting IP precedence in the real world?
Thanks Scott,
Clear and informative as usual! I still have a couple of further questions.
My question was partly an attempt to question the fact that IP precedence
does more than act as a tag when used to mark traffic. You can mark traffic
and act based on that. But, unlike a tag, the precedence value on its own
has some significance, even if you don't explicitly match the precedence.
So, if I were to mark traffic leaving my network and it wan't rewritten, I
*could* achieve some benefit if WFQ or WRED were employed by upstream
routers.
I see what you mean by perception. I though that there was a general scale
to IP precedence with a higher value meaning higher priority. Why not mark
traffic with a precedence of 5? Any device configured to look at precedence
will see it. In times of congestion, it may actually make a difference if
WFQ or WRED is configured, or they've configured the routers to treat
precedence 5 as a priority.
I also understand that there is no incentive to an ISP to treat my traffic
with a higher priority, but if the above were through, it would be possible
to use the features available in QoS features like WFQ and WRED on Cisco
routers to improve the performance of traffic without the explicit agreement
of my ISP.
This really is a theoretical question, because I don't have this type of
experience to date. I am just curious about how the real world treats
precedence, as opposed to private or labbed environments.
BTW, apart from WFQ and WRED is there any other QoS feature which uses
precedence by default.
Also, Wendells QoS book seems to be out of print. Anyone know where I can
buy a copy at a reasonable price?
Thanks
Roy
On Tue, 22 Feb 2005 12:11:49 -0500, Scott Morris wrote:
> Not so much as rewriting them (though some do) but more of ignoring
> the implications.
>
> The problem is that QoS is end-to-end, and what motivation does any
> ISP
have
> to treat your traffic better than one of their other customers?
>
> They'll be happy to sell you SLAs though. And then will tell you how
> to mark it.
>
> Beyond that though, we have problems in perception. What is important
> to your network may not be to mine. What makes a "prec 3" level of
importance
> to you may not have any correlation in my world.
>
> This is where the Per Hop Bevior RFC's came in to try to give a common
> set of interpretations at any network along the way. The Assured
> Forwarding
> (AF) classes and Expedited Forwarding (EF) class came from there.
>
> But it still boils town to money and motivation!
>
> HTH,
>
> Scott Morris, MCSE, CCDP, CCIE4 (R&S/ISP-Dial/Security/Service
> Provider) #4713, JNCIP, CCNA-WAN Switching, CCSP, Cable Communications
> Specialist,
IP
> Telephony Support Specialist, IP Telephony Design Specialist, CISSP
> CCSI #21903 swm@emanon.com
>
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf
> Of
Roy
> Dempsey
> Sent: Tuesday, February 22, 2005 11:56 AM
> To: Cisco certification
> Subject: Setting IP precedence in the real world?
>
> Hi all,
>
> I've always struggled with understanding and applying the correct QoS
> solution to a particular problem. I'm working hard at understanding
> when
and
> where different solutions are appropriate. As with other topics, I
> think I understand the theory, but need to understand the real-world
> application properly.
>
> While working through MQC, I couldn't help wondering about how
> important
ip
> precedence values are in the real world and, in particular, the
> Internet.
On
> private networks, I can see the value in setting the ip precedence
> value, and providing levels of service using these values.
>
> When sending traffic onto the Internet, is there any value in setting
> precedence values? Would ISPs rewrite these values as standard? Even
> if
ISPs
> don't explicitly use precedence values, wouldn't it be beneficial
> where
QoS
> devices like WRED and WFQ are employed, as these can use the
> precedence value?
>
> Thanks again,
> Roy
>
> ______________________________________________________________________
> _
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Thu Mar 03 2005 - 08:51:24 GMT-3