wiki:QinqResults

Version 1 (modified by jwilliams@bbn.com, 14 years ago) (diff)

migrated initial text from SysengWiki:QinQCapabilites

Purpose

This document outlines Exploring QinQ using various vendor's switch hardware (DUT - Device Under Test). This document covers:

  • initial QinQ overview
  • how to configure each DUT for QinQ
  • feasibility testing
  • interoperability testing

QinQ Overview

QinQ can be used to "tunnel" a particular VLAN of a "customer" network through a "service" network. This is a very important concept in GENI to allow multiple "customer" VLANS to be interconnected through the VLANS of regional service providers.

No image "QinQOverview.jpg" attached to QinqResults

In the image above VLAN A and VLAN B are two VLANs that span between "Network 1" and "Network 2". These networks can be at two separate sites. VLANs A & B are tunneled through the "Intermediate network" using VLAN Q. This allows the customer network VLAN usage to be independent of the Intermediate Network while still allowing the customer VLAN traffic to transverse the intermediate network. The Intermediate Network has a "separate VLAN ID space" than the two other network sites. VLAN X could be the same VLAN ID as either VLAN A or VLAN B without any collision.

The distinction between the different types of VLANs results in a discussion of the Ethernet Frame's EtherType field. An Ethernet Frame's header uses an EtherType field to specify what type of data is contained within the frame.

For VLANs the value in this field is, in large part, determined by the switch's port's configuration.

VLAN type EtherType value tag type port type note
none per protocol untagged frame access port IP = 0x800, ARP = 0x0806, etc
customer 0x8100 tagged frame 802.1q VLAN trunk port 802.1q VLAN tagging
service 0x8a88 QinQ tagged frame 802.1ad (QinQ) VLAN trunk port 8021.ad provider tunnel

For 802.1q VLANs and 802.ad QinQ VLANs the EtherType is actually a "Tag Protocol Identifier (TPID)" that, along with other tagging information, is inserted after the frame's source MAC address field. As this TPID is at the same byte offset as the original EtherType field, it is common to refer to this field as the EtherType field when discussing VLANs. The image below illustrates this distinction. Each tag adds 4 Bytes of data to the Ethernet frame.

No image "untaggedTaggedQinqTagged.jpg" attached to QinqResults

As the above illustration shows, the frame size was increased (by 4 Bytes) per each added tag. The size implications are shown in the table below.

use header size tag size MTU FCS total frame size
standard ethernet 14 0 1500 4 1518
802.1q VLAN trunk 14 +4 1500 4 1522
802.1ad (QinQ) VLAN "tunnel" 14 +4 +4 1500 4 1526

The 802.3ac standard increased the maximum frame length from 1518 to 1522 Bytes specifically, and exclusively, for VLAN tags. If a switch vendor adheres to this standard then VLAN tags can be viewed as part of the header (MTU stays as 1500) while QinQ tags require an increase of MTU (to 1504) to handle the inner tag. The test MTU's in QinQ specifically addresses the per-vendor MTU implementation.

Simulated Network Topology

A single network site's topology, as shown in the Overview diagram, is simulated for this test set up within a single switch to fully exercise the DUTs VLAN and QinQ capabilities. A single physical switch will be implementing both customer and service VLANs.

A VLAN trunk between a DUT's customer and service VLANs looks like a "jumper" cable -- a customer VLAN trunk must be created out of a customer-level switch trunk port and conencted to a service-level switch's acess port. This is a consequence of simulating multiple virtual switches within a single physical switch This jumper is shown below in the Simulated Network Topology diagram. For complete end-to-end testing Two DUT's, each configured as shown, would be connected using the respective QinQ port.

No image "SimulatedTestTopology.jpg" attached to QinqResults

Device Summary

Poblano NEC IP 8800 switch (DUT) -- see hwNecIP8800? for reference and usage
Habanero HP ProCurve 6600 switch (DUT) -- see hwHpProcurve6600? for reference and usage
Basil Cisco Catalyst 3750 switch (DUT)-- hwCiscoCatalyst3750? for reference and usage
Naboo VM Server
azzalle host on Naboo for testing
gotland host on Naboo for testing
lagnace host on Naboo for testing
skaldia host on Naboo for testing
wireshark host used for traffic monitoring (laptop)
wasabi Switch used for internal network connections

Test Topology with Internal Network

This section outlines the configuration steps necessary to integrate the DUTs into the BBN internal network to allow for testing and configuration.

No image "TestTopology.jpg" attached to QinqResults

The above diagram represents all major "classes" of connections that are between between physical devices. these connections are implemented as required per test.

To set up the test network the following steps were necessary:

  1. Configure VM test hosts.
  2. Configure each DUT.
    1. management IP addresses.
    2. management network connections.
    3. port and VLAN assignment.

Configuration

All VMs and DUTs have management IP addresses on the network 128.89.91.0/24, with physical connection via wasabi and/or naboo. This IP network is not used for any test traffic.

VM hosts

The following settings are from #506 and #537.

host DNS IP
azzalle azzalle.gpolab.bbn.com 128.89.91.9
gotland gotland.gpolab.bbn.com 128.89.91.10
lagnace lagnace.gpolab.bbn.com 128.89.91.11
skaldia skaldia.gpolab.bbn.com 128.89.91.12

Vm host's test network IP addresses and physical interfaces on Naboo.

host IP [eth1] -- NIC on naboo
azzalle 10.20.1.9 naboo[vmnic3] (2nd NIC on left card)
gotland 10.20.1.10 naboo[vmnic0] (3rd NIC on left card)
lagnace 10.20.1.11 naboo[vmnic1] (4th NIC on left card)
skaldia 10.20.1.12 :naboo[vmnic6] (the top NIC on the second card)

DUT's Port and VLAN Assignments

It is advantageous to use the same port assignments for each DUT to ensure consistency and prevent confusion. Each DUT will have the same port assignment as shown below. This configuration allows for a single configuration to accommodate all planned testing without reconfiguration between tests.

No image "PortstwoQinQtunnels.jpg" attached to QinqResults

Port Note
1 QinQ Port
2 management port
3 extra management port reserved for direct connection
4 host port
5 cvlan trunk - to port 6
6 svlan access - to port 5
7 cvlan trunk - to port 8
8 svlan access - to port 7
9 Not used
10 host port
11 host port
12 host port
13 cvlan trunk - to port 14
14 svlan access - to port 13
15 host port
16 Not used

Management IP Addresses

switch IP subnet VLAN management port wasabi's port
poblano 128.89.91.6/25 900 2 gi0/2
habanero 128.89.91.7/25 900 2 gi0/3
basil 128.89.91.8/25 900 2 gi0/4

See OpsNumbering for gateway, broadcast, etc.

Test Outline

This section outlines the various tests to perform on a DUT as well as between DUTs.

Unit Tests

These tests are performed on a single DUT.

QinQ Tagging

Purpose
Verify that a given DUT's QinQ port sends double-tagged QinQ frames in the expected format. For switches to understand that the trunking mechanism is a QinQ VLAN trunk the Ethernet's Header must contain the appropriate QinQ header field type indication (0x8a88).

No image "QinQTaggingTestOverview.jpg" attached to QinqResults

Method
Verify frames originating from the test host are tagged as appropriate using the Wireshark host.

  • HP: cvlans and svlans are used to distinguish port type. svlan trunk (QinQ) ports use the 0x88a8 value.
  • NEC: configures a QinQ trunk-port explicitly with the setting "switch dot1q ethertype 8a88" for a given port.
  • CISCO: The access port for the QinQ portion needs configured for QinQ, the QinQ trunk (ES) port is configured with 0x88a8.

QinQ with OpenFlow

Purpose
Verify that QinQ can operate within an OpenFlow enabled switch. This test verifies that a DUT can be configured to control QinQ VLANS while running the OpenFlow software. OpenFlow will not be configured to perform any flow-based traffic shaping.

Method
Enable OpenFlow and perform all experiments.

QinQ Tag Exclusivity

Purpose
This test insures that the customer VLAN ID and service VLAN ID ranges are mutually exclusive.

No image "QinQExclusitivityOverview.jpg" attached to QinqResults

Method
The DUT is configured with service VLAN 667 used for QinQ as well as a customer VLAN 667. If successful the Wireshark host should see a frame with outer tag ID 667 as well as inner tag 667.

VLAN and QinQ VLAN on Same Port

Purpose
This test explores the behavior of allowing a normal VLAN trunk and a service VLAN (QinQ) trunk to be allowed on the same port.

No image "CustomerAndServiceSamePort.jpg" attached to QinqResults

Method

  • VLAN 128 is a customer VLAN going out of port 1.
  • Service VLAN 667 tunneling customer VLAN 3702 will also be trunked on port 1.
  • Use wireshark to verify the appropriate frames are tagged for VLAN 128 or QinQ-tagged with outer VLAN 667 and inner VLAN 3702.

Integration Tests

QinQ between DUTs

Purpose
Verify that hosts in the same VLAN on opposite sides of a QinQ tunnel can communicate.

No image "QinQBetweenDUTs.jpg" attached to QinqResults

Method
Ping between hosts in the same tunneled VLAN.

MTU's in QinQ

Purpose
For QinQ to work efficiently, the ports trunking QinQ frames must accept allow for a frame size of 1526. A particular vendor's switch will require correct MTU configuration to prevent fragmentation.

According to 802.3ac frame size was increased to 1522 to allow a 1500 Byte MTU for VLAN tags. QinQ Trunks would require an MTU of 1504.

use header size MTU size total size switch MTU command note
standard ethernet 18 1500 1518 none required standard frame
802.1q VLAN trunk 18 + 4 1500 1522 none required effects header only
QinQ VLAN "tunnel" 18 + 4 1500 + 4 1526 system MTU 1504 MTU adjusted for inner tag

This ignores other MTU modification requirements such as:

  • MPLS VPN pass-through (two 4-byte labels)
  • Various Frame in Frame tunneling schemes (18 Bytes to 50 Bytes)

Method

ping 10.20.1.11 -M do -s "$((1500-20-8))" -c 1 > MTU_validation.txt
  • -M do: (return error if ping packet would fragment)
  • -s packetsize: size of payload:
    • 1500 = desired MTU
    • 20 = IP header size (would be IN the frame's MTU)
    • 8 = ICMP header size (would be IN the frame's MTU)

VLAN and QinQ VLAN on Same Port

Purpose
If this was feasible for two DUTs this test verifies that the hosts in this VLAN on separate DUTs can communicate.

No image "CustomerAndServiceSamePortQinQ.jpg" attached to QinqResults

Method
Simply ping between hosts in the same customer VLAN going out of the service port.

Latency: VLAN vs QinQ

Purpose
This test compares the best-case Round Trip latency of two hosts using a standard VLAN trunk and then a QinQ VLAN trunk.

VLAN Latency

No image "LatencyVlan.jpg" attached to QinqResults

QINQ Latency
As this test relies on end-to-end host connectivity over a QinQ tunnel, the setup is the same as "the test QinQ between DUTs". This diagram is included again here for completeness. No image "LatencyQinQ.jpg" attached to QinqResults

Method
This will be accomplished by using ping to report the round trip latency over 10 seconds. This test will ping the corresponding host once to "primes the queue" to prevent the ARP request from the first ping skewing the max, average and mdev values.

ping 10.20.1.11 -c 1; ping 10.20.1.11 -c 10 | tee aFile.txt

QinQ Traffic Simulations

Purpose
This test will involve inter-VLAN traffic tunneled across a QinQ Tunnel.

No image "TwoQinQTunnelsOneVlanPerTunnel.jpg" attached to QinqResults

Method

Use iperf to generate multiple TCP and UDP streams.

TCP
TCP streams allow for max throughput simulations.

iperf -c 10.20.1.11 -i 60 -t $(("60*60*8"))>aFile.txt &

server:
iperf -s -i 60 > aFile.txt &

UDP
As UDP doesn't have an ACK mechanism it is necessary to "prime the queue" to prevent the server from dropping any traffic due to fast-sender issues. This is necessary as iperf UDP server would see large amounts of dropped traffic until the ARP request resolved.

client:
ping -c 1 10.20.1.11; iperf -c 10.20.1.11 -u -i 1 -b 50M | tee aFile.txt

server:
iperf -s -u -i 1 | tee aFile.txt

Unit Testing

NEC

Configuration

VLAN port participation:

poblano# show vlan config
Date 2010/04/17 17:00:15 UTC
VLAN counts:12
ID   Name             Status  Ports
   1 VLAN0001         Down    -
 128 VLAN0128         Up      0/1,0/4
 667 VLAN0667         Up      0/1,0/6,0/13-15
 668 VLAN0668         Up      0/1,0/8
 900 VLAN0900         Up      0/2-3
3701 VLAN3701         Down    -
3702 VLAN3702         Up      0/5,0/11
3703 VLAN3703         Up      0/7,0/12
3704 VLAN3704         Up      0/5,0/10

VLAN port membership with trunking configuration. Only relevant interface info is shown, for complete configuration see the NEC Appendix.

interface gigabitethernet 0/1
  switchport dot1q ethertype 88a8
  switchport mode trunk
  switchport trunk allowed vlan 128,667-668
!
interface gigabitethernet 0/2
  switchport mode access
  switchport access vlan 900
!
interface gigabitethernet 0/3
  switchport mode access
  switchport access vlan 900
!
interface gigabitethernet 0/4
  switchport mode access
  switchport access vlan 128
!
interface gigabitethernet 0/5
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 3702,3704
!
interface gigabitethernet 0/6
  switchport mode dot1q-tunnel
  switchport access vlan 667
!
interface gigabitethernet 0/7
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 3703
!
interface gigabitethernet 0/8
  switchport mode dot1q-tunnel
  switchport access vlan 668
!
interface gigabitethernet 0/10
  switchport mode access
  switchport access vlan 3704
!
interface gigabitethernet 0/11
  switchport mode access
  switchport access vlan 3702
!
interface gigabitethernet 0/12
  switchport mode access
  switchport access vlan 3703
!
interface gigabitethernet 0/13
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 667
!
interface gigabitethernet 0/14
  switchport mode dot1q-tunnel
  switchport access vlan 667
!
interface gigabitethernet 0/15
  switchport mode access
  switchport access vlan 667

QinQ Tagging

the NEC correctly tagged the frames for QinQ transmission, as shown below.

Wireshark sees QinQ double-tagged frame 667:3702. (e.g. 667 is the outer vlan, 3702 is the "wrapped" vlan)

Ethernet frame
    type: 802.1ad Provider Bridge (QinQ) (0x88a8)
    IEEEE 802.1ad ID:667
        ID667
        type 802.1Q virtual LAN (0x8100) ID 3702
            ID 3702
            IP (0x08000)
                PAYLOAD

No image "NEC_QinQ_inner_3702_outer_667.jpg" attached to QinqResults

QinQ with OpenFlow

All QinQ testing was conducted while Poblano was running OpenFlow 0.9 firmware; no flows were active. Regardless of which Firmware is used to boot the device (original or OpenFlow) the start-up configuration is retained.

Ports and VLANs can either be used by OpenFlow or as part of the production network. Future tests will explore QinQ with multiple active flows.

Customer VLAN out QinQ Port

128 VLAN0128         Up      0/1,0/4 

VLAN 128's ports were configured as follows:

interface gigabitethernet 0/1
  switchport dot1q ethertype 88a8
  switchport mode trunk
  switchport trunk allowed vlan 128,667-668
!
interface gigabitethernet 0/4
  switchport mode access
  switchport access vlan 128

No image "NEC_VLAN128NoQinQ.jpg" attached to QinqResults

VLAN 128 is capable of being sent out port 1, however it's tagged type is "0x8a88" (service VLAN), "not 0x8100" (customer VLAN). This implies that the Switch on the other side of the trunk must be a service VLAN; sending the VLAN as a "normal" VLAN isn't possible in this configuration.

Same Inner and Outer VLAN Tags

 667 VLAN0667         Up      0/1,0/6,0/13-15

From VLAN 667's port participation it's appears that there's no distinction between customer and service VLANs despite various ports being configured to tag for QinQ vs normal tagging. If all the ports are indeed in the same VLAN the VLAN trunk (aka "jumper) from ports13<->14 should create a broadcast storm which would be observable on port 1 (STP is disabled).

1. connecte azzalle to Nec's port 15
2. Wireshark connected to NEC port 1.
3  On azzalle: 10.20.1.11 -c 1 

No image "NEC_SameIds_broadcast_storm.jpg" attached to QinqResults A broadcast storm was induced from a single ping packet. The image also shows the continual nesting of VLAN headers as the frame continues to loop between access and trunk ports. From this it doesn't look possible to tunnel the same customer and service VLAN using one switch. This is an artifact caused by trying to emulate, on a single physical switch, service and customer VLANS of the with the same VLAN ID bridged with an Ethernet cable.


HP

Overview

The HP needs explicit configuration to operate using both customer and service VLANS. This configuration is applied as the preparatory global step of assigning mixed vlan mode.

Configuration

NOTE: The HP Procurve manual states that the switch only supports 2048 VLANS (half of the usual 4096).

Note: Make sure the command below is the first configuration step. Setting or changing this value causes the switch to immediately reboot – and completely wipe its running configuration during the process – to take effect.

qinq mixedvlan

Ports used in svlans are not allowed to participate in the GARP VLAN Registration Protocol (GVRP). The switch prompts when an incorrect assignment is attempted and entering the following command fixes the problem:

int 1 unknown-vlans disable

For this reason, VLAN 667 (in the test example) was configured as an svlan and 3702 as a vlan.

habanero# show vlan

 Status and Counters - VLAN Information

  Maximum VLANs to support : 2000                 
  Primary VLAN : DEFAULT_VLAN
  Management VLAN :            

 VLAN ID Name                 Type  | Status     Voice Jumbo
  ------- -------------------- ----- + ---------- ----- -----
  1       DEFAULT_VLAN         cvlan | Port-based No    No   
  128     VLAN128              svlan | Port-based No    No   
  667     VLAN667              svlan | Port-based No    No   
  668     VLAN668              svlan | Port-based No    No   
  900     VLAN900              cvlan | Port-based No    No   
  3702    VLAN3702             cvlan | Port-based No    No   
  3703    VLAN3703             cvlan | Port-based No    No   
  3704    VLAN3704             cvlan | Port-based No    No   
vlan 1 
   name "DEFAULT_VLAN" 
   untagged 9,13-48,49-50,51-52 
   no untagged 1-8,10-12 
   no ip address 
   exit 
vlan 3702 
   name "VLAN3702" 
   untagged 11 
   tagged 5 
   no ip address 
   exit 
vlan 3703 
   name "VLAN3703" 
   untagged 12 
   tagged 7 
   no ip address 
   exit 
vlan 3704 
   name "VLAN3704" 
   untagged 10 
   tagged 5 
   no ip address 
   exit 
vlan 900 
   name "VLAN900" 
   untagged 2-3 
   ip address 128.89.91.7 255.255.255.128 
   exit 
qinq mixedvlan
svlan 128
   name "VLAN128"
   tagged 1
   untagged 4
   exit
svlan 667
   name "VLAN667"
   tagged 1
   untagged 6
   exit
svlan 668
   name "VLAN668"
   tagged 1
   untagged 8
   exit

QinQ Tagging

The HP setup worked similarly to the NEC. However to specify a VLAN as used for QinQ the VLAN must be marked as a service VLAN. Note the HP needs to be configured to support mixed VLANs (see HP appendix) before this can be configured - changing this setting removes all VLAN configuration from the switch.

Wireshark sees QinQ double tagged frame 667:2702 (e.g. 667 is the outer vlan, 3702 is the "wrapped" vlan). Wireshark reports the correct QinQ frame header type

Ethernet frame
    type: 802.1ad Provider Bridge (QinQ) (0x88a8)
    IEEEE 802.1ad ID:667
        ID667
        type 802.1Q virtual LAN (0x8100) ID 3702
            ID 3702
            IP (0x08000)
                PAYLOAD

No image "HP QinQ tags.jpg" attached to QinqResults

QinQ with OpenFlow

All QinQ testing was conducted while Habanero was running OpenFlow 0.9 firmware; no flows were active. Regardless of which Firmware is used to boot the device (original or OpenFlow) the start-up configuration is retained.

Ports and VLANs can either be used by OpenFlow or as part of the production network. Future tests will explore QinQ with multiple active flows.

Customer VLAN Over QinQ Port

Creating VLAN 128 as a customer VLAN did not work; adding port 1 to the VLAN's participating ports failed:

habanero(vlan-128)# tagged 1  
Ports 1 will lose their svlan memberships.
Do you want to continue? [y/n] n

Clearly, port 1 should remain a service port to stay a QinQ port - the answer was no.

However, deleting the customer vlan 128 and then creating a service VLAN 128 worked:

habanero(config)# svlan 128
habanero(svlan-128)# interface svlan 128
habanero(svlan-128)# untagged 4
Interfaces that are GVRP enabled cannot be members of svlans.
Use the interface level 'unknown-vlans' command to disable port gvrp.
habanero(svlan-128)# exit     
habanero(config)# interface 4 unknown-vlans disable 
habanero(config)# interface svlan 128
habanero(svlan-128)# untagged 4
Ports 4 will lose their cvlan memberships.
Do you want to continue? [y/n] y
habanero(svlan-128)# show vlan 128
habanero(svlan-128)# tagged 1
habanero(svlan-128)# show vlans 128

 Status and Counters - VLAN Information - VLAN 128

  VLAN ID : 128    
  Name : VLAN128             
  Type : svlan
  Status : Port-based
  Voice : No 
  Jumbo : No 

  Port Information Mode     Unknown VLAN Status    
  ---------------- -------- ------------ ----------
  1                Tagged   Disable      Up        
  4                Untagged Disable      Down      

This behavior seems consistent with the NEC. It is possible to send a non-tunneled VLANs out a service trunk port. However, those VLANs would be service VLANs, not customer VLANs.

Same Inner and Outer VLAN Tags

Tried to create cvlan of 667 fails as there's already a svlan with 667.

habanero(config)# vlan 667
VLAN type mismatch. VID 667 is of type 'svlan'.

This was not as expected; HP's distinction between customer and service VLANs seems to imply (besides the implicit type tagging on trunk ports) that this configuration would be possible. Though surprising, this behavior is consistent with the NEC.


Cisco

Overview

The cisco 3750 requires the SFP ES module for QinQ operation. installed as well as the appropriate licensing. See http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps5532/prod_qas09186a00801eb822.html Note that all configuration in this section refers to "port 1" This refers to this ES port (GigabitEthernet?1/1/1 ). Ticket #533 outlines the procedure for getting the Cisco configured properly.

Configuration

Trunk Negotiation

To allow for QinQ the Cisco Discovery protocol (CDP) should be disabled per dot1q (normal) VLAN tunk port.

interface FastEthernet1/0/6
 switchport access vlan 667
 switchport mode dot1q-tunnel
 no cdp enable                  <--HERE
!         
interface FastEthernet1/0/8
 switchport access vlan 668
 switchport mode dot1q-tunnel
 no cdp enable                    <--HERE
MTU

The standard system mtu does not change ES port configuration. To adjust the MTU on these ports use the system mtu jumbo command to allow for QinQ tagging.

    3750(config)# system mtu jumbo 9000
    3750(config)# exit
    3750# reload
QinQ access ports

The QinQ Access ports (which connect to the standard dotq trunk ports) are configured to by using `switchport mode dot1q-tunnel.

 interface FastEthernet1/0/6
 switchport access vlan 667
 switchport mode dot1q-tunnel
 no cdp enable
QinQ Trunk ports

The QinQ trunk ports are set to participate in the same VLANs as the QinQ access ports. The trunk ethertype is set to the QinQ type: 0x88A8

 interface GigabitEthernet1/1/1
 switchport trunk allowed vlan 128,667,668
 switchport mode trunk
 switchport nonegotiate
 switchport trunk dot1q ethertype 88A8
 speed auto 1000

QinQ Tagging

Wireshark sees QinQ double tagged frame 667:2702 (e.g. 667 is the outer vlan, 3702 is the "wrapped" vlan). Wireshark reports the correct QinQ frame header type

Ethernet frame
    type: 802.1ad Provider Bridge (QinQ) (0x88a8)
    IEEEE 802.1ad ID:667
        ID667
        type 802.1Q virtual LAN (0x8100) ID 3702
            ID 3702
            IP (0x08000)
                PAYLOAD

QinQ with OpenFlow

None - Cisco doesn't support OpenFlow firmware.

Customer VLAN Out QinQ Port

VLAN 128 is capable of being sent out ES port 1, however it's tagged type is "0x8a88" (service VLAN), "not 0x8100" (customer VLAN). This implies that the Switch on the other side of the trunk must be a service VLAN; sending the VLAN as a "normal" VLAN isn't possible in this configuration. This is the same behavior as the NEC switch.

Same Inner and Outer VLAN Tags

Simular to NEC configuration ports with identical inner and outer VLAN IDs, when connected together via a jumer, cause a broadcast storm.


Interoperability Testing

NEC<->HP

QinQ Between DUTs

Ping from azzalle to gotland succeeded.

MTU's in QinQ

QinQ between NEC and HP can transmit an MTU of 1500 without fragmentation.

Results:

ping 10.20.1.11 -M do -s "$((1500-20-8))" -c 1 > MTU_validation.txt
1480 bytes from 10.20.1.11: icmp_seq=1 ttl=64 time=0.434 ms

while adding 1 byte gave:

ping 10.20.1.11 -M do -s "$((1500-20-8+1))" -c 1
From 10.20.1.9 icmp_seq=1 Frag needed and DF set (mtu = 1500)

NOTE Poblano's current configuration for QinQ , due to previous experimental configuration, has a QinQ MTU of 1508 and a VLAN Trunk MTU of 1504 - This configuration is different than 802.3ac, but may be correct for this vendor. Habanero doesn't currently have any explicit MTU settings. Setting Poblano's QinQ MTU to 1504 and the VLAN trunk MTU to 1500 results in no "Echo" nor "Destination Unreachable (Fragmentation required, and DF flag set)" responses. This seems to imply that the ICMP Echo Request successfully made it to the destination host, but the ICMP Echo Response was dropped when leaving the HP's QinQ port. More investigation is required.

See the NEC IP8800 Manual: Configuration Settings, Vol. 3, section 1.4.3 for more information.

VLAN and QinQ VLAN on Same Port

After configuring both the NEC's and HP's VLAN 128 to be a service VLAN (as outlined in the Unit test sections) end-to-end communication was possible.

Latency: VLAN vs QinQ

Ping across the tested ports. The first "ping -c 1" "primes the queue" to prevent the ARP request from the first ping skewing the max, average and mdev values.

ping 10.20.1.11 -c 1; ping 10.20.1.11 -c 10 | tee aFile.txt

VLAN Trunk Only

--- 10.20.1.11 ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9066ms
rtt min/avg/max/mdev = 0.192/0.239/0.315/0.036 ms

QinQ Trunk

--- 10.20.1.11 ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9072ms
rtt min/avg/max/mdev = 0.196/0.244/0.284/0.038 ms

With all the extra hops for QinQ, an added RTL of 0.244-0.239=0.005ms seems more than reasonable. Given the short length of cabling this is a close approximation to "switching delay". Again, this was an Ideal baseline; there was no other traffic on the link for this baseline measurement.

QinQ Traffic Simulations

TCP
Naboo's VM hosts were capping out at ~430 Mbps for TCP traffic (+/- 7Mbps based on quick scanning of my iperf log files per 10sec over 10 minutes) This is a limitation of Naboo (VM server) and is not a limitation of any DUTs. This was with only 1 pair communicating - full 1Gb capacity was available. Testing both pairs over QinQ still resulted in transmission of ~430Mbps per pair (logged every minute over 8 hours). I noticed no downward performance trend - but again I am currently eyeballing. With two end-to-end pairs, we're still under the max capacity of the link.

iperf -c 10.20.1.11 -i 60 -t $(("60*60*8"))>file.txt &

UDP
UDP testing revealed minimal packet-loss and jitter was 0.023 ms ± 0.002ms.

ping -c 1 10.20.1.11; iperf -c 10.20.1.11 -u -i 1 -b 50M

NEC<->CISCO

QinQ Between DUTs

Ping from azzalle to gotland succeeded.

MTU's in QinQ

QinQ between NEC and Cisco can transmit an MTU of 1500 without fragmentation.

Results:

ping 10.20.1.11 -M do -s "$((1500-20-8))" -c 1 > MTU_validation.txt
1480 bytes from 10.20.1.11: icmp_seq=1 ttl=64 time=0.434 ms

while adding 1 byte gave:

ping 10.20.1.11 -M do -s "$((1500-20-8+1))" -c 1
From 10.20.1.9 icmp_seq=1 Frag needed and DF set (mtu = 1500)

VLAN and QinQ VLAN on same port

After configuring both the NEC's and Cisco's VLAN 128 to be a service VLAN (as outlined in the Unit test sections) end-to-end communication was possible.

Latency: VLAN vs QinQ

Was not tested

QinQ Traffic Simulations

TCP
Naboo's VM hosts were capping out at ~230 Mbps for TCP traffic (+/- 7Mbps based on quick scanning of my iperf log files per 10sec over 10 minutes) This is a limitation of Naboo (VM server) and is not a limitation of any DUTs. This was with only 1 pair communicating - full 1Gb capacity was available. This bandwidth is less than the amount possible when the NEC<->HOP tests were conducted. However, there are quite a few more VMs on Naboo the now.

UDP
UDP testing indicated no packet-loss.


HP<->CISCO

QinQ Between DUTs

MTU's in QinQ

QinQ between NEC and HP can transmit an MTU of 1500 without fragmentation.

Results:

ping 10.20.1.11 -M do -s "$((1500-20-8))" -c 1 > MTU_validation.txt
1480 bytes from 10.20.1.11: icmp_seq=1 ttl=64 time=0.434 ms

while adding 1 byte gave:

ping 10.20.1.11 -M do -s "$((1500-20-8+1))" -c 1
From 10.20.1.9 icmp_seq=1 Frag needed and DF set (mtu = 1500)

See the NEC IP8800 Manual: Configuration Settings, Vol. 3, section 1.4.3 for more information.

VLAN and QinQ VLAN on same port

After configuring both the HP's and Cisco's VLAN 128 to be a service VLAN (as outlined in the Unit test sections) end-to-end communication was possible.

Latency: VLAN vs QinQ

Was not tested

QinQ Traffic Simulations

TCP
Naboo's VM hosts were capping out at ~230 Mbps for TCP traffic (+/- 7Mbps based on quick scanning of my iperf log files per 10sec over 10 minutes) This is a limitation of Naboo (VM server) and is not a limitation of any DUTs. This was with only 1 pair communicating - full 1Gb capacity was available. This bandwidth is less than the amount possible when the NEC<->HOP tests were conducted. However, there are quite a few more VMs on Naboo the now.

UDP
UDP testing indicated no packet-loss.


Appendix

NEC

See hwNecIP8800? for reference and usage.

Configuration

poblano# show running-config 
#Last modified by operator at Sat Apr 17 17:58:25 2010 with version 11.1.C
!
hostname "poblano"
!
ip host poblano 128.89.91.6
!
ip domain name bbn.com
!
ip name-server 128.33.0.20
!
vlan 1
  name "VLAN0001"
!
vlan 22
  name "BBN OpenFlow 1"
!
vlan 23
  name "BBN OpenFlow 2"
!
vlan 24
  name "BBN OpenFlow Control Vlan"
!
vlan 128
!
vlan 667
!
vlan 668
!
vlan 900
!
vlan 3701
!
vlan 3702
!
vlan 3703
!
vlan 3704
!
spanning-tree disable
spanning-tree mode pvst
!
interface gigabitethernet 0/1
  media-type rj45
  mtu 1508
  switchport dot1q ethertype 88a8
  switchport mode trunk
  switchport trunk allowed vlan 128,667-668
!
interface gigabitethernet 0/2
  media-type rj45
  switchport mode access
  switchport access vlan 900
!
interface gigabitethernet 0/3
  media-type rj45
  switchport mode access
  switchport access vlan 900
!
interface gigabitethernet 0/4
  media-type rj45
  switchport mode access
  switchport access vlan 128
!
interface gigabitethernet 0/5
  mtu 1504
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 3702,3704
!
interface gigabitethernet 0/6
  mtu 1504
  switchport mode dot1q-tunnel
  switchport access vlan 667
!
interface gigabitethernet 0/7
  mtu 1504
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 3703
!
interface gigabitethernet 0/8
  mtu 1504
  switchport mode dot1q-tunnel
  switchport access vlan 668
!
interface gigabitethernet 0/9
  switchport mode dot1q-tunnel
  switchport access vlan 22
!
interface gigabitethernet 0/10
  switchport mode access
  switchport access vlan 3704
!
interface gigabitethernet 0/11
  switchport mode access
  switchport access vlan 3702
!
interface gigabitethernet 0/12
  switchport mode access
  switchport access vlan 3703
!
interface gigabitethernet 0/13
  mtu 1504
  switchport dot1q ethertype 8100
  switchport mode trunk
  switchport trunk allowed vlan 667
!
interface gigabitethernet 0/14
  switchport mode dot1q-tunnel
  switchport access vlan 667
!
interface gigabitethernet 0/15
  switchport mode access
  switchport access vlan 667
!
interface gigabitethernet 0/16
  switchport mode dot1q-tunnel
  switchport access vlan 23
!
interface gigabitethernet 0/17
  switchport mode trunk
  switchport trunk allowed vlan 22-23
!
interface gigabitethernet 0/18
  switchport mode access
  switchport access vlan 24
!
interface gigabitethernet 0/19
  switchport mode access
!
interface gigabitethernet 0/20
  switchport mode access
!
interface gigabitethernet 0/21
  switchport mode access
!
interface gigabitethernet 0/22
  switchport mode access
!
interface gigabitethernet 0/23
  switchport mode access
!
interface gigabitethernet 0/24
  switchport mode access
!
interface tengigabitethernet 0/25
  switchport mode access
!
interface tengigabitethernet 0/26
  switchport mode access
!
interface vlan 1
!
interface vlan 24
  ip address 171.67.74.60 255.255.255.240
  no ip proxy-arp
!
interface vlan 900
  ip address 128.89.91.6 255.255.255.128
!
ip route 0.0.0.0 0.0.0.0 128.89.91.1
!
line vty 0 2
!
ftp-server
!
ntp server 192.1.100.189
ntp server 192.1.249.10
!
poblano#   

Useful Commands

The following are some notes taken while learning the NEC switch syntax: See hwNecIP8800? for more discussion.

Getting started

  • login: operator

Administrative commands Enable mode (necessary to do just about anything and doesn’t prompt for a password)

enable	

enter configuration mode

configure	Enter configuration mode
(When making configuration changes, the console prints a "!" character to indicate there are unsaved changes...)

Password management

clear password <username>  #Clear the user's password.  
                                                     #The password utility does not allow setting a NULL password
                                                     #use this command to clear it
password <username>	     #Change a user's password - defaults to the currently logged-in user

config (mode)

save	Save current configuration – look for the “!” characters!
vlan [vlan number]	Activate the specified VLAN.  No parameters assigned, just the entry is available to be used elsewhere...
interface gig eth 0/1	Select the interface.  It's really "interface gigabitethernet 0/1" but it allows abbreviations when they are non-ambiguous...
interface vlan [vlan id]	Select the vlan to configure.  The only (useful) thing I've found is the ability to associate a VLAN with an IP address and THAT was only useful to define the HOME interface of the device...
media [rj45/sfp]	While configuring an interface.  Ports 1-4 are dual-option, RJ-45 is ethernet, SFP is the fiber port.  One or the other is enabled...
switch dot1q ethertype 8a88	Switches the ethertype announced between 802/1ad (“tunnel”) and 802.1q mode on this interface.

HP

See hwHpProcurve6600? for reference and usage.

Current Configuration

NOTE the actual password for the HP has been replaced with XXXXX for display. If you intend on using this output as a configuration you must replace XXXXX with the appropriate password.

habanero# show running-config 

Running configuration:

; J9452A Configuration Editor; Created on release #K.14.53o

hostname "habanero" 
max-vlans 2000 
module 2 type J94yyA 
module 3 type J94zzA 
module 5 type J94wwA 
module 6 type J94wwA 
no stack 
interface 1 
   unknown-vlans Disable 
exit
interface 4 
   unknown-vlans Disable 
exit
interface 6 
   unknown-vlans Disable 
exit
interface 8 
   unknown-vlans Disable 
exit
ip default-gateway 128.89.91.1 
vlan 1 
   name "DEFAULT_VLAN" 
   untagged 9,13-48,49-50,51-52 
   no untagged 1-8,10-12 
   no ip address 
   exit 
vlan 3702 
   name "VLAN3702" 
   untagged 11 
   tagged 5 
   no ip address 
   jumbo 
   exit 
vlan 3703 
   name "VLAN3703" 
   untagged 12 
   tagged 7 
   no ip address 
   jumbo 
   exit 
vlan 3704 
   name "VLAN3704" 
   untagged 10 
   tagged 5 
   no ip address 
   jumbo 
   exit 
vlan 900 
   name "VLAN900" 
   untagged 2-3 
   ip address 128.89.91.7 255.255.255.128 
   exit 
qinq mixedvlan
svlan 128
   name "VLAN128"
   tagged 1
   untagged 4
   exit
svlan 667
   name "VLAN667"
   tagged 1
   untagged 6
   exit
svlan 668
   name "VLAN668"
   tagged 1
   untagged 8
   exit
jumbo ip-mtu 1508
jumbo max-frame-size 1526
sntp server priority 1 192.1.249.10 3
ip ssh filetransfer
snmp-server community "public" unrestricted
oobm
   ip address dhcp-bootp
   exit
no tftp client
no tftp server
no autorun
password XXXXX

Useful Commands

HP ProCurve 6600 Useful Commands The following are some notes taken while learning the HP switch syntax: (no login username) Enters "setup" screen to set things like the name (habanero), IP address and netmask:

setup 	
	System Name: habanero
	Default Gateway: 128.89.72.1
	IP Config [Manual]
	Spanning Tree Enabled [No]
	IP Address: 128.89.72.141
	Subnet Mask: 255.255.254.0
	"Save" saves and exits...

show relevant info

show [stuff]

IP address information

show ip	

VLAN information

show vlan	

VLANs Can Be Tagged and Untagged Resets the configuration to a default state - make sure the current config is backed up before doing it!)

qinq mixedmode	

Other VLAN commands vlan refers to customer VLANs, svlan refers to service VLANs

vlan 3702 untagged 37	        #Causes port 37 to stop tagging VLAN 3702 traffic (undoes a “vlan tag” command)
svlan 667 tagged 38	                #Causes port 38 to tags traffic as SVLAN 667 (turns it into a tunneling-trunk port)
int 45 unknown-vlans disable	#Disables GVRP [needed for SVLAN membership]
int 45 unknown-vlans learn	#Enables GVRP
int 46 qinq port-type customer-network
vlan 1 no tagged 37,45,46	Removes the ports from VLAN 1 if they were configured as untagged participants
vlan 1 no untagged 37,45,46	Removes the ports from VLAN 1 if they were configured as tagging participants

Cisco

See hwCiscoCatalyst3750? for reference and usage.

Configuration

basil#show running-config 
Building configuration...

Current configuration : 2772 bytes
!
version 12.2
no service pad
service timestamps debug uptime
service timestamps log uptime
no service password-encryption
!
hostname basil
!
enable secret 5 $1$m1O6$lT/GyoO4dZOw0bvD9j/wH/
enable password operator
!
no aaa new-model
system mtu routing 1500
ip subnet-zero
!
vtp mode transparent
!
no file verify auto
!
spanning-tree mode pvst
spanning-tree extend system-id
no spanning-tree vlan 1,128,667-668,900,3702-3704
!         
!
!
vlan internal allocation policy ascending
!
vlan 128,667-668,900,3702-3704 
!
!
interface FastEthernet1/0/1
!
interface FastEthernet1/0/2
 switchport access vlan 900
 switchport mode access
!
interface FastEthernet1/0/3
 switchport access vlan 900
 switchport mode access
!
interface FastEthernet1/0/4
 switchport access vlan 128
 switchport mode access
!
interface FastEthernet1/0/5
 switchport trunk encapsulation dot1q
 switchport trunk allowed vlan 3702,3704
 switchport mode trunk
 switchport nonegotiate
!
interface FastEthernet1/0/6
 switchport access vlan 667
 switchport mode dot1q-tunnel
 no cdp enable
!
interface FastEthernet1/0/7
 switchport trunk encapsulation dot1q
 switchport trunk allowed vlan 3703
 switchport mode trunk
 switchport nonegotiate
!
interface FastEthernet1/0/8
 switchport access vlan 668
 switchport mode dot1q-tunnel
 no cdp enable
!
interface FastEthernet1/0/9
!
interface FastEthernet1/0/10
 switchport access vlan 3704
 switchport mode access
!
interface FastEthernet1/0/11
 switchport access vlan 3702
 switchport mode access
!
interface FastEthernet1/0/12
 switchport access vlan 3703
 switchport mode access
!
interface FastEthernet1/0/13
 switchport trunk encapsulation dot1q
 switchport trunk allowed vlan 667
 switchport mode trunk
 switchport nonegotiate
!
interface FastEthernet1/0/14
 switchport access vlan 667
 switchport mode dot1q-tunnel
 no cdp enable
!
interface FastEthernet1/0/15
 switchport access vlan 667
 switchport mode access
!
interface FastEthernet1/0/16
!
interface FastEthernet1/0/17
!
interface FastEthernet1/0/18
!
interface FastEthernet1/0/19
!
interface FastEthernet1/0/20
!
interface FastEthernet1/0/21
!
interface FastEthernet1/0/22
!
interface FastEthernet1/0/23
!
interface FastEthernet1/0/24
!
interface GigabitEthernet1/0/1
!
interface GigabitEthernet1/0/2
!
interface GigabitEthernet1/1/1
 switchport trunk allowed vlan 128,667,668
 switchport mode trunk
 switchport nonegotiate
 switchport trunk dot1q ethertype 88A8
 speed auto 1000
 no cdp enable
!
interface GigabitEthernet1/1/2
 speed auto 1000
!
interface Vlan1
 no ip address
!
interface Vlan900
 ip address 128.89.91.8 255.255.255.128
!
ip classless
ip http server
!
!
!
!
!
control-plane
!
!
line con 0
 exec-timeout 0 0
line vty 0 4
 password operator
 login
line vty 5 15
 password operator
 login
!
end

Useful Commands


Related Wiki Pages

  • OpsSwitchConfig - guide for configuring switches for use in lab network
  • OpsNumbering - Various configuration paramters, IP dresses, gateways, etc.

Related Tickets

  • #333 - Initial NEC (and later HP) configuration.
  • #533 - Initial exploration of QinQ support for Cisco switch.
  • #501 - QinQ report for QinQ capabilities of NEC switch (depends on this and the performance testing). Outlines Report.
  • #535 - QinQ interoperability testing (this document).
  • #539 - Install NEC OpenFlow Firmware.
  • #456 - Install HP OpenFlow Firmware.
  • #506 - Internal network configuration for switch management IPs and host VMS for testing.
  • #537 - Configuring switches for internal network management access.