[c-nsp] how to validate cell rates?

MADMAN david.madland at qwest.com
Thu Jan 20 14:48:55 EST 2005



Gert Doering wrote:

> Hi,
> 
> On Thu, Jan 20, 2005 at 09:50:07AM -0700, Clinton Work wrote:
> 
>>Gert, I agree with your calculation, but it really depends upon how the 
>>Telco has built the ATM VC inside their network. The Telco could have
>>used a VBR PVC with an unforgiving CDVT (Cell Delay Variance Tollerance) 
>>value. 
> 
> 
> This is my suspicion, yes.  They *claim* it's UBR, and UBR is what I have
> configured...
> 
> Are the CDVT values specific to the configured VC types?  They claim
> a CDVT value of "750", though I'm not sure what to make out of it, or
> how to match this to Cisco config statements.
> 
> 
>>The only soltuion may be reducing your PCR value until the packet 
>>loss goes away.
> 
> 
> This is what we did now - "ubr 2150" seems to make everything well-behaving,
> but it means we're paying for bandwidth we can't use, and their own
> documentation how we should configure things is not working, and I'm 
> not going to accept this so easily.

     Depends.  If the problem is the telco yes you may be loosing 
bandwidth.  If it is the problem I have encountered then no you are not.
> 
> (The return path is even worse.  Their own Ethernet->ATM/SDSL bridge is
> not doing *any* shaping, so as soon as the customer side is sending 
> large amounts of traffic, the switch polices, both AAL5 and OAM cells,
> and we see packet loss *and* "line protocol down" flaps.  Their official
> recommendation is now "connect a router that can do traffic shaping".
> We've been through this with their ADSL/Ethernet product, and they
> *have* fixed their CPEs, it just took half a year... *sigh*)

   Yes you don't want to connect a nonshaping device to the public ATM. 
You get great thruput but lousy "goodput"!

   Dave
> 
> gert

-- 
David Madland
CCIE# 2016
Sr. Network Engineer
Qwest Communications
612-664-3367

"Emotion should reflect reason not guide it"


More information about the cisco-nsp mailing list