1. Procket’s IPv6 Implementation
16-Nov-18

2. IPv6 Features Available Now (Release 2.3)
Wire-rate performance
IPv6 Routing Protocols
BGP4+
OSPFv3
RIPng
IS-IS
IPv6 Transition Mechanisms
Dual stack
IPv6 over GRE
6 to 4 Tunnels (RFC 3056)
IPv6 Services
SSH, SCP, Telnet, FTP & DNS
Ping and Traceroute
IPv6 Addressing
Neighbor Discovery and Auto-Configuration
16-Nov-18

3. IPv6 Roadmap Release 2.4 IPv6 Multicast Phase I Security
PIM-SM SSM
Security
IPv6 Packet filtering PCLs
128-bit source address
128-bit destination address
TCP/UDP source port
TCP/UDP destination port
Next Header
ICMP type & code
TCP flags
Hop limit
System Management
IPv6 MIBs
TCP IPv6 MIB (RFC 2452)
UDP IPv6 MIB (RFC 2454)
Textual Conventions and General Group IPv6 MIB (RFC 2465)
ICMPv6 Group MIB (RFC 2466)
16-Nov-18

4. IPv6 Roadmap Release 2.5 IPv6 Multicast Phase II System Management
MBGP
PIM-SM
Bi-Dir
ASM
Multicast Listener Discovery (MLDv1 & v2)
RFC draft v2
System Management
IPv6 Multicast Routing MIB (enterprise)
OSPFv3 MIB (draft-ietf-ospf-ospfv3-mib-06.txt)
IPv6 BGP4 MIB (enterprise)
IPv6 PIM MIB (enterprise)
In addition, SSM translation is supported for IPv4 in 2.4; and may be for IPv6 in This allows you to have an SSM cloud in your core even if the host don’t support IGMPv3 (MLDv2 if IPv6).
Windows XP supports IGMPv3 but not MLDv2…only major operating system that supports MLDv2 is FreeBSD.
16-Nov-18

5. Internal IPv6 Protocol Test Results
Use Tahi and Spirent for IPv6 conformance testing
Perform interoperability testing with Cisco and Juniper
Successfully tested the following:
BGP4+
4M paths (or 300 peers with 12k-14k IPv6 prefixes each)
300 peers
OSPFv3
15,000 IPv6 routes
300 neighbors
ISIS for IPv6
200 adjacencies
Max forwarding table entries
Max static routes
14,000 IPv6 routes
BGP peer capacity is for both IPv4 and IPv6 combined since it’s an integrated protocol.
16-Nov-18

6. Internal IPv6 Protocol Test Results OSPFv3
Perform interoperability testing with Cisco and Juniper
Use Anvil and Spirent for conformance testing
Use IXIA and Spirent for emulation
In-house developed tests: ~250
Tested with 300 neighbors while redistributing 15k IPv6 routes
Tested with 200 neighbors and 20*20 grid (total of ~2000 LSAs)
Convergence Test Results:
Test Description
Convergence Time*
Time to receive 10k redistributed routes from single neighbor:
68 sec
Time to receive 20k redistributed routes from single neighbor:
84 sec
Time to receive LSAs for 20*20 grid from single neighbor:
50 sec
Time to receive LSAs for 40*40 grid from single neighbor:
55 sec
Conformance testing is to test for conformance to standard
Emulation is stress and scalability testing (tools are topology generators)
For convergence tests – Procket router is connected to IXIA; tester issues a ‘no shutdown’ and starts the clock when the interface comes up (as shown in debug mode).
* Measured as time from interface up to routes installed in hardware.
16-Nov-18

7. Internal IPv6 Protocol Test Results IS-IS for IPv6
Perform interoperability testing with Cisco and Juniper
Use QARobot for IPv4 conformance testing
Currently no conformance test for IPv6
Use IXIA and Spirent for emulation
In-house developed tests: ~100
Tested with 200 adjacencies while redistributing 15k IPv6 routes
Tested with 200 adjacencies and 20*20 grid (total of ~2000 LSPs)
Convergence Test Results:
Test Description
Convergence Time*
Time to receive 10k redistributed routes over single adjacency:
50 sec
Time to receive 20k redistributed routes over single adjacency:
60 sec
Time to receive LSPs for 20*20 grid over single adjacency:
1 min 35 sec
Time to receive LSPs for 40*40 grid over single adjacency:
2 min 30 sec
Conformance testing is to test for conformance to standard
Emulation is stress and scalability testing (tools are topology generators)
The fact that there is currently no conformance test for IPv6 is not a significant risk since there are only 2 new TLVs.
* Measured as time from interface up to routes installed in hardware.
16-Nov-18

8. Internal IPv6 Protocol Test Results BGP4+
Perform interoperability testing with Cisco and Juniper
Use QARobot for IPv4 conformance testing
Currently no conformance test for IPv6
Use IXIA and Spirent for emulation
In-house developed tests: ~200
Time to establish 300 peers: ~ 25 sec
Time to establish 300 peers and receive 20k routes from each: ~17 min
Conformance testing is to test for conformance to standard
Emulation is stress and scalability testing (tools are topology generators)
16-Nov-18

9. Internal IPv6 Performance Test Results 10G Ethernet
Media 10G Ethernet
Platform PRO/8812
Type 100% load, bi-directional
Duration 180 seconds
Software Version 2.2
Date 6/23/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
100%
100%
100%
100%
Frame Size
(bytes)
Min Latency
(µsec)
Avg Latency
Max Latency
Throughput 70
23.98
24.38
25.12
100%
128
24.72
24.96
25.50
256
25.20
25.46
25.94
512
26.08
26.34
27.08
1518
29.32
29.70
30.44
3000
33.84
34.11
34.68
6000
42.16
42.43
43.08
9166
50.80
51.19
51.94
24.38
24.96
25.46
26.34
29.70
34.11
42.43
51.19
Test tool required 70 byte minimum packet size for latency tests. 70
128
256
512
1518
3000
6000
9166
16-Nov-18

10. Internal IPv6 Performance Test Results OC192c PPP
Media OC192c (PPP)
Platform PRO/8812
Type 100% load, bi-directional
Duration 180 seconds
Software Version 2.2
Date 6/23/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
100%
100%
100%
Frame Size
(bytes)
Min Latency
(µsec)
Avg Latency
Max Latency
Throughput 64
12.41
25.73
27.18
100%
128
23.45
26.92
27.75
256
22.07
27.09
27.81
512
26.12
28.49
1508
24.96
30.12
30.83
4476
33.25
38.47
39.45
6000
36.55
42.80
43.57
9156
Frame size not supported by test tool
42.80
38.47
26.92
27.09
27.75
30.12
25.73 64
128
256
512
1508
4476
6000
9156
16-Nov-18

11. Internal IPv6 Performance Test Results OC48c PPP
Media OC48c (PPP)
Platform PRO/8812
Type 100% load, bi-directional
Duration 180 seconds
Software Version 2.2
Date 6/23/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
100%
100%
100%
100%
Frame Size
(bytes)
Min Latency
(µsec)
Avg Latency
Max Latency
Throughput 64
28.68
34.05
57.24
100%
128
29.48
36.79
48.18
256
30.14
46.07
151.34
512
33.12
38.99
48.58
1508
38.14
40.35
43.24
4476
58.76
61.08
64.88
6000
67.60
70.65
73.72
9156
81.50
88.39
91.16
88.39
70.65
61.08
46.07
38.99
40.35
34.05
36.79 64
128
256
512
1508
4476
6000
9156
16-Nov-18

12. Internal IPv6 Performance Test Results 1G Ethernet
Media 1G Ethernet
Platform PRO/8812
Type 100% load, bi-directional
Duration 180 seconds
Software Version 2.2
Date 6/23/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
100%
100%
100%
100%
Frame Size
(bytes)
Min Latency
(µsec)
Avg Latency
Max Latency
Throughput 70
26.34
26.49
27.62
100%
128
27.74
27.98
28.36
256
30.02
30.34
30.94
512
34.54
34.99
35.54
1518
53.52
54.00
55.08
3000
70.08
70.48
70.92
6000
101.26
101.82
102.54
9166
128.90
129.35
129.90
129.35
101.82
70.48
54.00
34.99
30.34
26.49
27.98
Test tool required 70 byte minimum packet size for latency tests. 70
128
256
512
1518
3000
6000
9166
16-Nov-18

13. Spirent IPv6 Performance Test Results
Tests performed in September 2003
At Spirent’s Calabasas, California facility
PRO/8812 High-Availability Router running PRO/1 software v2.2
Test equipment
Spirent TeraRouter Tester version 3.0 with TeraMetrics 1G and 10G cards
Fully Loaded PRO/8812
240 Ports of 1000BaseSX
24 Ports of 10GBaseLX
Single packet size and IMIX tests
IMIX was 2:1:1 ratio of 76, 576, and 1500 byte packets
Traffic fully meshed
Each port had a flow to every other 1G and 10G port
Very important notes:
These tests were designed to allow a third party (Spirent) to test a fully loaded chassis. This chassis had 240 ports of 1G Ethernet (called the ‘edge’) and 24 ports of 10G Ethernet (called the ‘core’).
There was no attempt to optimize the tests for latency or try and produce a ‘speed run’. These tests were meant to see what the chassis would do under fully loaded conditions with IPv6 traffic.
16-Nov-18

14. Spirent IPv6 Performance Test Results IPv6 Flows from 10G to 1G interfaces
Media 10G Ethernet to 1G Ethernet
Platform PRO/8812
Type 100% offered load, fully meshed
Duration 30 seconds
Software Version 2.2
Date 9/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
Frame Size
(bytes)
Avg Latency
(µsec)
Median Latency
Throughput 76
255.25
203.32
100%
576
1500
IMIX
880.70
249.01
For each test run (76, 576, 1500, IMIX) there are two latency figures. The AVERAGE (or MEAN) latency is the sum of all of the average latencies reported by the test tool for 10G to 1G traffic flows, divided by the total number of flows. This produces an ‘average of an average’.
The MEDIAN latency is the statistical representation of the middle of the latency numbers. The MEDIAN is often used to give a representation of how skewed the figures are from lowest to highest.
The MEDIAN latency for the IMIX is very different from the AVERAGE because of the wide range of values when using the different packet sizes.
The latency numbers on this slide will be very high compared to the next slide, because the traffic flows being measured here are subject to the queuing done when the egress is a 1G port. That is, because the ingress is faster than the egress, these figures will be higher than those on the next slide.
880.70
255.25 76
576
1500
IMIX
16-Nov-18

15. Spirent IPv6 Performance Test Results IPv6 Flows from 1G to 10G interfaces
Media 1G Ethernet to 10G Ethernet
Platform PRO/8812
Type 100% offered load, fully meshed
Duration 30 seconds
Software Version 2.2
Date 9/2003
Throughput(%) and average latency(µsec)
100%
100%
100%
100%
Frame Size
(bytes)
Avg Latency
(µsec)
Median Latency
Throughput 76
24.30
28.06
100%
576
34.85
38.39
1500
51.10
55.71
IMIX
47.12
51.30
51.10
47.12
34.85
24.30 76
576
1500
IMIX
16-Nov-18

16. 16-Nov-18