Basic BGP Convergence Benchmarking

1
Basic BGP Convergence Benchmarking

• Rajiv Papneja
• Bhavani Parise

2
Background and Current Status

• Terminology for Benchmarking BGP Device
  Convergence in the Control Plane (RFC 4098) /
  June 2005
• Within BMWG charter, and milestone to send the
  document for AD review in July 2010
• Team formed to produce the methodology draft
• draft-papneja-bmwg-bgp-convergence-meth-00.txt
• Rajiv Papneja, Bhavani Parise, Rodney Dunn, Mohan
  Nanduri. Jay Karthik, Eric Brendel

3
Scope of the Draft

• Data plane convergence with emphasis on BGP IPv4
• Scope limited to basic BGP convergence
• Considers convergence for BGP speakers core,
  border edge, and possibly route reflectors
• Considers BGP Failure/Convergence Events
• Considers dependencies on factors impacting
  convergence - such as number of peers, number of
  routes/peers, BGP/IGP interaction etc.
• Data Traffic characterization offered load
• Various test cases to account for iBGP, eBGP,
  failure convergence events
• Topologies Several 2 node, and 3 node setups

4
Test Methodologies

• RIB-IN convergence
• RIB-OUT convergence
• iBGP convergence
• eBGP convergence
• eBGP multi-hop convergence
• BGP Failure/Convergence Events
• Physical Link failure (local/remote)
• ECMP link failure (local/remote)
• BGP session failure
• Hard Reset tests - BGP Non-Recovering Hard Reset
• BGP Soft Reset
• BGP Route Withdrawal Convergence measurement
• BGP Path Attribute Change Convergence Time
• BGP Graceful Restart Convergence Time

5
Next Steps

• Publish initial version of the Methodology
  document by April 2010.
• Welcoming contributors and input from SP
  community
• Feedback welcome from the group on the test cases
• Question for the group in terms of preference
• Control vs. Data Plane BGP convergence
• Our initial approach is for the data plane
  convergence benchmarking