(no subject)

Hey Rick,

first i would like to Thanks a tons to u.

I have understood by ur above mail,still i will make things 100% clear,

 The topology

 CE-AA
---DS/PE-----CoreP/PE----ISP/PE------ISP/P--------ISP/PE----------- CE-BB.

FOR OUTBOUND TRAFFIC:

Many customers are connected to my distribution switch acting as a PE,the
link between DS nd CORE is 10G so what am thinking is not to classify or
mark the traffic on Distribution so that customers can use upto 10G till
core.On core i will apply bandwidth command and also shaping if needed on
each customer VRF interface according to SLA for bandwidth with customer by
the command service policy output.As u are aware it is a back to back VRF
with ISP.

Correct me if any thing wrong on above para.

FOR INBOUND TRAFFIC:

The traffic which i will recieve by the ISP on CORE i mean to say another
end customers (CE-BB),i will police that traffic according to customer SLA
for bandwidth.Correct me if am wrong.

From ur above mail as u said

The IP Precedence received from the another end customers through ISP
(INBOUND traffic to CORE) it will be just copied in EXP field in the MPLS
header and will be carried towards the distribution switch and then to
customers.

Thanks,for ur immediate response.

On Mon, Sep 28, 2009 at 12:59 AM, Rick Mur <rmur_at_ipexpert.com> wrote:

> Hi Adam,
> I'll try to answer your questions as I don't really understand them for
> 100% :-)
>
> The first one about back-to-back VRF is totally correct. All packets send
> in a back-to-back VRF configuration are just plain IPv4 packets.
>
> What do you mean by other clients? What traffic is traversing the link that
> you want to police? I don't what the bandwidth of that link is, but if you
> have 'clients' coming in over it, than of course you could set policing so
> that traffic will not exceed a certain rate. Indeed you could use the
> 'bandwidth' command and use CBWFQ, another option would be to use Generic
> Traffic Shaping. Remember that CBWFQ only works when the link is congested
> and traffic shaping almost always works (depending on your settings of
> course). Inbound policing is a good method of limiting how much a user can
> use, I would really communicate this to your clients as traffic will be
> dropped and you could get angry customers, so by informing them or
> mentioning it in a contract would be the best as they can then take their
> measures against it (like traffic-shaping on their side).
>
> By default nothing is changed in the packet as of IP precedence values that
> are already set. If you trust your customers to set the correct values by
> themselves, you could just leave it as is. If you don't or have other uses
> for it, further in the network than I would set them myself. By default NO
> QoS is applied when you don't configure a policy on an interface, default
> the IP precedence bits in the DSCP field are copied to the EXP field in the
> MPLS header.
>
> Like you said, by default the IP Prec bits are copied to the EXP field and
> relayed all the way until the MPLS header pops off. If the EXP value is
> changed along the way, by default this NEW value is copied to the IP packet
> again. You could disable this or specify a new value for it on the PE's.
>
> Either case I don't have a full view of the exact network topology or what
> you are planning to do. I hope I could give you satisfying answers :-)
>
> --
>
> Regards,
>
> Rick Mur
> CCIE2 #21946 (R&S / Service Provider)
> Sr. Support Engineer  IPexpert, Inc.
> URL: http://www.IPexpert.com <http://www.ipexpert.com/>
>
> On 27 sep 2009, at 20:54, adam gibs wrote:
>
> Hi,
>
> Can anybody help me for the below Queries of MPLS QOS
>
> Immediate response will be highly appreciated
>
> On Sun, Sep 27, 2009 at 12:16 PM, adam gibs <adamgibs7_at_gmail.com> wrote:
>
>> Rick,
>>
>> In back-to-back VRF the new label will be assigned on core towards
>> distribution because in back-to back vrf the packet forwarded to isp is
an
>> IP packet and packet which i will recieve from ISP will also be an IP
packet
>> because view from my end to ISP is as such as CE device communicating
>> through BGP and from ISP end he will thing that am as a CE device, Correct
>> me if am wrong,
>>
>> Suppose if am not policing inbound traffic on each VRF interface (coming
>> from ISP to CORE) it will utilize the link as much as he wants???? and
that
>> traffic will continue utilizing till distribution switch. The LINK between
>> the distribution and the core is 10G no such problem for this link but
what
>> about other clients which will also flow their traffic on link between ISP
>> and CORE they will survive,so what am thinking is the QOS what should be
>> implement on core for each VRF interface is 1 service policy output with
the
>> bandwidth command according the SLA with the customer and 1 service input
>> policy configured with the police command where exceed traffic will be
>> dropped,Correct me what am suggesting is write????
>>
>> Question 3
>>
>> The IP precedence what i will receive from ISP end i should alter on links
>> between distribution and core ?????? OR core should just relay the TOS
bits
>> to EXP bits,And which MPLS QOS will be applied on inbound traffic when i
>> dont configure an service policy output towards distribution switch,
>>
>> Question 4
>> Bydefault ip precedence are copied to EXP and they are just relayed to
>> other end of customer untill and unless any alteration has been done in
>> intermediate routers of ISP????????
>>
>> Immediate response will be appreciated.
>>
>> On Sun, Sep 27, 2009 at 1:03 AM, Rick Mur <rmur_at_ipexpert.com> wrote:
>>
>>> Hi Adam,
>>>
>>> To answer your first question it's definitely regular QoS as you don't
>>> exchange any labeled packets with your ISP, so there is no MPLS header to
be
>>> altered.
>>>
>>> The second question depends on how much the 10G link is utilized between
>>> the core and distribution. I would say that QoS would not be very much of
>>> use as most QoS techniques are based on congested links. It would only
have
>>> any use if you would implement policing to restrict certain types of
>>> traffic. Again on a 10G link, I would say that's not really necessary, but
I
>>> don't have an overview on how much traffic is flowing over that link :-)
>>> The topology between Core and Distribution is also quite strange. You
>>> terminate MPLS VPNs from ISP to DS? Or are it just VRF sub-interfaces
>>> (VRF-Lite)?
>>>
>>>
>>> --
>>>
>>> Regards,
>>>
>>> Rick Mur
>>> CCIE2 #21946 (R&S / Service Provider)
>>> Sr. Support Engineer  IPexpert, Inc.
>>> URL: http://www.IPexpert.com <http://www.ipexpert.com/>
>>>
>>>
>>> On 26 sep 2009, at 22:32, adam gibs wrote:
>>>
>>> The Connection is as below
>>>>
>>>> CE---DS/PE-----CoreP/PE----ISP/PE,
>>>>
>>>> The link between distribution and the core is fiber 10 gig i dont want
>>>> to
>>>> implement any bandwidth limitation on the links between distribution and
>>>> the
>>>> core i want let the customer use the 10 gig link upto the core but on
>>>> core
>>>> VRF sub-interfaces facing to ISP which QOS i have to apply traditional
>>>> QOS
>>>> or MPLS QOS ????.The packet will be a IPV4 am having a back to back vrf
>>>> connection with ISP.
>>>>
>>>> question 2:
>>>> The traffic which i will recieve from ISP definately i will police that
>>>> traffic according to CLient demand bandwidth but when it will travel
>>>> towards
>>>> the distribution switch i should implement QOS towards the distribution
>>>> switch???? or i shld implement MPLS QOS or i shld leave that traffic as
>>>> default,because i want the 10 Gig link between core and distribution
>>>> shld be
>>>> utilize as much as can.Can u advise me what are the drawback for this.
>>>>
>>>>
>>>> Blogs and organic groups at http://www.ccie.net
>>>>
>>>> _______________________________________________________________________
>>>> Subscription information may be found at:
>>>> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Mon Sep 28 2009 - 21:20:11 ART