IETF BMWG Work Items

1
IETF BMWG Work Items

• 65th IETF Meeting
• Dallas, TX
• Tuesday 3/21/06

2

• BENCHMARKING NETWORK LAYER
• TRAFFIC CONTROL MECHANISMS
• draft-ietf-bmwg-dsmterm-11.txt
• draft-ietf-bmwg-dsmmeth-01.txt
• Co-authors are
• Scott Poretsky of Reef Point, Jerry Perser of
  Veriwave, Shobha Erramilli of Qnetworx, and Sumit
  Khurana of Telcordia

65th IETF Meeting Dallas
3
Terminology

• draft-ietf-bmwg-dsmterm-12.txt, Terminology for
  Benchmarking Network Layer Traffic Control
  Mechanisms
• Terminology completed WGLC
• Required Co-Chair review prior to IESG revealed a
  few issues that are now corrected
• Clarify that delay is Forwarding Delay
• Minor grammar and format issues
• Only remaining outstanding issue is reference to
  Jitter definition in obsoleted EF PHB RFC
• Ready for IESG review?

4
Methodology

• draft-ietf-bmwg-dsmmeth-01.txt, Methodology for
  Benchmarking Network Layer Traffic Control
  Mechanisms
• Applies many of the terms from the Terminology
  draft
• Test Cases
• Undifferentiated Response
• Traffic Control Baseline Performance
• Traffic Control Performance with Forwarding
  Congestion

5
Methodology Baseline Test Cases

• Undifferentiated Response
• This is the baseline case with
• Multiple flows of SA/DA pairs and DSCP0 (BE)
• Aggregate Offered Load is lt Forwarding Capacity
• Traffic Control Baseline Performance
• This is the DSCP baseline case with
• Multiple flows of SA/DA pairs
• Multiple DSCP values
• Aggregate Offered Load is lt Forwarding Capacity

• Expected
• Vector
  
• \/
  
• --------- Offered Vector
  ---------
• lt----------------------------
  ----
• 
  
• DUT
  Tester
• 
  
• gt
  
• Output Vector
  
• ---------
  ---------

6
Methodology Congestion Test Cases

• Expected
• Vector
  
• \/
  
• --------- Offered Vector
  ---------
• lt----------------------------
  ----
• lt----------------------------
  ----
• DUT
  Tester
• 
  
• gt
  
• Output Vector
  
• ---------
  ---------

• Traffic Control Performance with Forwarding
  Congestion
• This is the DSCP congestion case with Link
  Congestion
• This is the DSCP baseline case with
• Multiple flows of SA/DA pairs
• Multiple DSCP values
• Aggregate Offered Load is gt Forwarding Capacity
• ADD Test Case
• Traffic Control Performance with DSCP Congestion
• No Link Congestion, but configured DSCP Bandwidth
  is Exceeded

• Any input from WG?
• Other test cases to add?
• Any comments for methodology?

7

• BENCHMARKING IGP DATA PLANE
• ROUTE CONVERGENCE
• draft-ietf-bmwg-igp-dataplane-conv-app-10.txt
• draft-ietf-bmwg-igp-dataplane-conv-term-10.txt
• draft-ietf-bmwg-igp-dataplane-conv-meth-10.txt
• Co-authors are
• Scott Poretsky of Reef Point and
• Brent Imhoff of Juniper Networks

65th IETF Meeting Dallas
8
Current Status

• draft-ietf-bmwg-igp-dataplane-conv-app-10.txt,
  Considerations for Benchmarking IGP Data Plane
  Route Convergence
• draft-ietf-bmwg-igp-dataplane-conv-term-10.txt,
  Terminology for Benchmarking IGP Data Plane Route
  Convergence
• draft-ietf-bmwg-igp-dataplane-conv-meth-10.txt,
  Benchmarking Methodology for IGP Data Plane Route
  Convergence
• -08 successfully completed 2nd WGLC
• -09 issued to correct IETF NITs and incorporate
  comments for formatting and clarification from Al
  Morton, Thomas Eriksson, and Timmons Player
• -10 incorporates comments from Cross-Area
  Reviewer, Sue Hares (last step for IESG review)

9
Changes for 09

• Clean-up Normative/Informative References
• Clarify time measurement granularity is to
  milliseconds
• Specify the packet size includes Payload, IP
  header, and Link-Layer header
• Clarify last sentence of Convergence Packet loss
  discussion
• Fix figures with formatting error of the Tester
• Change "this draft describes" to "this document
  describes"
• Make Consistent use of term Throughput (not
  Forwarding Rate)
• Found rfc3978 Section 5.4 paragraph 1 boilerplate
  (on line 696), which is fine, but also found
  rfc2026 Section 10.4C paragraph 1 boilerplate on
  line 42. It should be removed.
• Considerations (Applicability) missing form feeds
• Some lines
• between 73 to 77 characters long (26 instances)
• with control characters (52 instances)
• With an extra space between words (5 instances)

10
Cross Area Review

• Overall comment - very well done! Document is
  accurate and well thought out.
• A few document edits/nits found and fixed in 10
• One comment not incorporated
• It would be very good to replicate the equations
  used by cisco for ISIS or IGP convergence as an
  appendix
• LoC(p) D O QSP (h F) SPF(n) RIB (p)
  FIB(p) DD CRR
•      D link outage  
•       0 Originate OSPF
•       QSP queue the ls updates
•       HF hops by flooding time
•       SPF(n) SPF calculation time
•       RIB(p) Routing RIB update time
•       FIB(p) - FIB update time
•       DD - Logical circuit update time
•       CRR Recursive Lookup for BGP
• That equation, while being very useful, does not
  fit directly into this IGP work.  It includes
  parameters that are White Box measurements, BGP
  time, and factors for multiple hops.  Since it
  was suggested to be in an appendix I felt more
  comfortable excluding it from this single box,
  black box, IGP benchmark. 

11
Next Steps

• -10 Ready for IESG Review?

12

• BENCHMARKING NETWORK DEVICES
• UNDER ACCLERATED STRESS
• draft-ietf-bmwg-acc-bench-term-08.txt
• draft-ietf-bmwg-acc-bench-meth-04.txt
• (draft-ietf-bmwg-acc-bench-meth-ebgp-00.txt
• draft-ietf-bmwg-acc-bench-meth-opsec-00.txt)
• Co-authors are
• Scott Poretsky of Reef Point and
• Shankar Rao of Qwest

65th IETF Meeting Dallas
13
Current Status

• Terminology
• draft-ietf-bmwg-acc-bench-term-07.txt,
  Terminology for Accelerated Stress Benchmarking
• -08 changes incorporate action items from IETF 64
• Specified the benchmark Recovery Time in
  micro-second resolution
• Added discussion that benchmarks span multiple
  dimensions and each can be compared as the
  methodology user requires for the DUT
  application.
• Renamed "degraded forwarding rate to "forwarding
  rate degradation"
• General Methodology
• draft-ietf-bmwg-acc-bench-meth-05.txt,
  Methodology Guidelines for Accelerated Stress
  Benchmarking
• -05 will incorporate action items from IETF 64.
• To be submitted by end of April.

14
Next Steps

• Is Terminology ready for WGLC?
• -05 Methodology will incorporate comments from
  IETF 64 and BMWG mailing list. To be posted by
  end of April.

15

• Backup Slides

16
Example Stress Test Configuration Set
Control Plane 30 BGP Peers (2 EBGP, 28
IBGP) 28 OSPF Adjacencies 400K route
instances 175K routes in FIB MPLS
Disabled Multicast Protocols Disabled 16K IPsec
Tunnels 32K IPsec SAs 16K IKE SAs IPsec SA
Lifetime 8 hours IKEv2 SA Lifetime 8
hours DPD Disabled
Security Plane 100K Stateful Firewall
Sessions 64K Firewall Rules DOS-Protection
Enabled Management Plane 20 SSH Sessions 4
RADIUS Servers with round-robin Logging
enabled SysLog enabled Statistics
enabled Data Plane Interfaces qty 4
GigE Data Rate 4 Gbps Packet Size 1500
bytes QoS Disabled
17
Example Stress Test Test Conditions

• Startup Conditions (as configured on Tester)
• BGP and OSPF pre-configured and negotiation
  starts immediately
• 50 IPsec Tunnels established per second
• 1500 Stateful Firewall Sessions established per
  second
•  
• Instability Conditions (as configured on Tester)
• 1 Interface Shut/No Shut per minute
• 1 OSPF Interface Cost Change per hour
• 100 IPsec Tunnels flapped (setup/teardown) per
  second
• 20 IKEv2/IPsec Rekeys per second
• RADIUS Server lost every 30 minutes
• Continuous DOS Attacks (using Nessus)
• Close/Open 1 SSH session per minute
• Enter SHOW, Config, and Errored commands for
  every open session
• 1 SNMP GET per second
• 1 FTP File Transer of 100Mb every second
• Tester is Test Device or System of Test Devices

18
Example Stress Test Benchmarks

• DEVICE 1
• 1. Configuration Sets achieved
• 2. Startup Phase Benchmarks
• Stable Aggregate forwarding Rate 4Gbps
• Stable Latency 110 usec
• Stable Session Count
• 30 BGP Peers
• 28 OSPF Adjacencies
• 16K IPsec Tunnels
• 3. Apply Instability Conditions
• 4. Instability Phase Benchmarks
• Unstable Aggregate Forwarding Rate 3.5Gbps
• Degraded Aggregate Forwarding Rate 0.5Gbps
• Unstable Latency 110usec
  
• Unstable Uncontrolled Sessions Lost 126

• DEVICE 2
• 1. Configuration Sets achieved
• 2. Startup Phase Benchmarks
• Stable Aggregate forwarding Rate 4Gbps
• Stable Latency 150 usec
• Stable Session Count
• 30 BGP Peers
• 28 OSPF Adjacencies
• 16K IPsec Tunnels
• 3. Apply Instability Conditions
• 4. Instability Phase Benchmarks
• Unstable Aggregate Forwarding Rate3.3Gbps
• Degraded Aggregate Forwarding Rate 0.7Gbps
• Unstable Latency 170usec
  
• Unstable Uncontrolled Sessions Lost 4000

• Configuration Set in this test was reduced from a
  previous test because Device 2 crashed at 20
  hours
• Test was repeated with 3rd Configuration Set to
  obtain a Recovery Time for Device 2