FW: RE: [bmwg] Updates to MPLS Protection Meth/Term draft- Request forcomments/review

I'd like to comment on the questions posed below.

Using accurate packet timestamping mechanisms it is feasible to measure data plane convergence much more accurately, modern measurement tools have timestamp resolution down to 10ns resolution. When using packet timestamps, the resolution and accuracy of a convergence or switchover measurement will be mostly limited by the packet frequency on the link. 1,000,000 packets per sec will provide a resolution of 1us for a data plane convergence measurement if packet timestamps are used, requiring about 50% of a GbE link's available bandwidth. 

I believe 1ms is an appropriate sampling interval for data plane switchover and convergence tests. This is feasible in most scenarios, and better matches the high availability expectations of modern services.

Regards,
Cary Wright

________________________________

*	To: <bmwg at ietf.org <mailto:bmwg@DOMAIN.HIDDEN> > 
*	Subject: RE: [bmwg] Updates to MPLS Protection Meth/Term draft- Request forcomments/review 
*	From: "Samir Vapiwala \(svapiwal\)" <svapiwal at cisco.com <mailto:svapiwal@DOMAIN.HIDDEN> > 
*	Date: Thu, 8 Mar 2007 09:28:28 -0500 

Further to Rajiv's earlier email we as MPLS Protection benchmarking Author team would like to seek input to the following unaddressed items on the mailing list so that we can get closure on these open items prior to the BMWG meeting:

 Since the packet smapling interval recommended in the "draft-ietf-bmwg-igp-dataplane-conv-meth-12.txt,  section 3.2.5 Convergence Time Metrics" may not be applicable to the MPLS FRR scenario where failover times have to be below 45ms, here are some of our thoughts or questions to the list: 

1.	We would like to adapt similar approach the IGP convergence as this defiinition applies to current work item 
2.	We feel that 100ms is large for MPLS FRR application, under which the target is always keeping the failover times below 45ms. 
3.	We are considering changing this number to 10ms, and would expect test tool to offer this smapling interval

Please provide your input to this suggested value so that we can have a conclusion before the next meeting.

Regards,

MPLS Protection Benchmarking Author Team