Re: [nsp] Getting real throughput on ATM PVC´s

[Wayne (cisco nsp)]

The behavior in the referenced doc should be accurate.  Measuring PVCs on
GSRs w/ 12.0(17)ST2 and 7500s w/ 12.0(11)S3 both exhibit this behavior (not
including the cell header and aal5 padding).  For a maxed-out VBR PVC (i.e.
stabilizing on a xmission rate of the SCR), you can see this by polling the
PVC utiliz off the ATM switch you will see the actual cell rate.

For example, if using MRTG to graph the BW utiliz on a VBR PVC w/ a 10Mbps
SCR that was "full", the PVC graph for the router would show a flat looking
graph at around 8.5Mbps.  The PVC measured via the ATM switch would show a
flat graph of around 10Mbps (i.e. the router shaped the cell rate on the ckt
to the SCR)  This is the expected behavior to prevent any ATM switch in the
path from policing the traffic.

HTH,

-Wayne Gustavus
  CCIE #7426



----- Original Message -----
From: "Adenilson Raniery Sarges Pontes" <raniery at rnp.br>
To: "Cisco-NSP" <cisco-nsp at puck.nether.net>
Sent: Friday, February 07, 2003 8:09 AM
Subject: [nsp] Getting real throughput on ATM PVC´s


>
> Dear all,
>
>  I´d like to get your experiences on measuring throughput on ATM PVCs
> using SNMP. Basically, we have GSR 12008 routers with IOS 12.0(21)S2, and
> Quad-port OC3c ATM linecards. PVCs were configured (one for each
> subinterface), with SCRs ranging from 4Mbps up to 38 Mbps.
>
> My question is: what snmp counters do you use to get *real* throughput on
> ATM PVCs?
>
> According to some Cisco docs, e.g
>
http://www.cisco.com/en/US/tech/tk39/tk51/technologies_tech_note09186a008009
3c9a.shtml
>
> subinterface MIB counters _ifInOctets and ifOutOctets_ do not include
> (deterministic) 5-byte cell header and (variable) AAL5 padding.
>
> Counters on CISCO-AAL5-MIB _cAal5VccInOctets and cAal5VccOutOctets_ seem
> to include AAL5 padding but I´m not sure, since some statements in the doc
> above sounded dubious to me :(
> Also, these are 32-bits counters leading to wrap problems with higher
> capacity PVCs.
>
> We´re looking for better measurements (or at least good estimatives) of
> our internal traffic on PVCs. Nowadays, we´ve been using subinterface
> counters but measured traffic is (obviously) lower than real traffic
> injected on telcos links.
>
> Any help would be really appreciated.
>
> Regards
>
> Raniery Pontes
> :::::::::::::::::::::::::
> Network Operations Center (NOC)
> National Research Network - Brazil
> http://www.rnp.br/index_en.html
>
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> http://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/