Software Defined Networking using VXLAN - Linux Foundation

IBM Presentation Template Full Version

Route Short Circuit (RSC)

Host A

# ping 21.0.0.A ip r add default via 20

20.0.0.A
54:A:20:0:0:A

Console VM A

NIC 192.168.1

Neighbor & FDB Host A:
ARP: 20.0.0.254 → 54:A:20:0:0:FE
21.0.0.B → 54:B:21:0:0:B
FDB: 54:B:21:0:0:FE → 1.2.3.4 router
54:B:21:0:0:B → 192.168.100.B

1) Look up router IP to MAC mapping in neighbor
2) Router flag set

a) Remote IP address in FDB entry igno
b) Look up destination IP address to MA
3) Replace destination MAC in inner header 54:A
4) Look up destination MAC in FDB and transmit

Traffic flow between VM A ↔ VM B

26 So

.0.0.254 Host B

ip r add default via 21.0.0.254

VMA VMB Virtual Bridge
Vxlan
A 21.0.0.B
54:B:21:0:0:B
100.A
Switch NIC 192.168.100.B

(1)
(2b)
r (2a)
(4)

r table

ored
AC mapping in neighbor table
A:20:0:0:FE → 54:B:21:0:0:1

to VTEP

oftware Defined Networking using VXLAN, Thomas Richter ([email protected]), LinuxCon 2013

IBM Presentation Template Full Version

Route Short Circuit 2 (Migration VM

# ping 21.0.0.A ip r add default via

20.0.0.A
54:A:20:0:0:A

Console VM A

Neighbor: F
20.0.0.254 → 54:0:1:2:3:4 (1)
21.0.0.254 → 54:0:1:2:3:4
20.0.0.A → 54:A:20:0:0:A
21.0.0.B → 54:B:21:0:0:B (3)

1) Look up router IP to MAC mapping in neighbo

2) Router flag set
a) Remote IP address in FDB entry igno
b) Look up destination IP address to MA

3) Replace destination MAC in inner header 54:A

4) Look up destination MAC in FDB and feed bac

Traffic flow between VM A ↔ VM B So

27

M A to Host B)

Host B

a 20.0.0.254 ip r add default via 21.0.0.254

VMA VMB Virtual Bridge
Vxlan
A 21.0.0.B
54:B:21:0:0:B
Switch
NIC 192.168.100.B

FDB:

54:0:1:2:3:4 → 1.2.3.4 router (2)
54:B:21:0:0:B → 0.0.0.0 (4)
54:A:20:0.0:A → 0.0.0.0

or table

ored
AC mapping in neighbor table
A:20:0:0:FE → 54:B:21:0:0:1
ck to local bridge (destination IP 0.0.0.0)

oftware Defined Networking using VXLAN, Thomas Richter ([email protected]), LinuxCon 2013

IBM Presentation Template Full Version

Open Stack Integration

See paper R. Cohen (References Nr 4)
– Map bridge name to VNI

Nova Quantum Agent Agent
Mgr Mgr Plugin Manager

Horizon Dashboard Agent Ext

Openstack control node

28 So

Nova Agent Agent
Agent VIF

Openstack compute node

Nova Agent Agent
Agent VIF

Openstack compute node

oftware Defined Networking using VXLAN, Thomas Richter ([email protected]), LinuxCon 2013

VXLAN Standard (2) Intranet/In

9. Ping REQ 20.0.0.8 Host 6

ff:ff:ff:ff:ff:ff 54:1:14:0:0:6 ARP REQ
1, 8 Who has 20.0.0.8 Tell 20.0.0.6

2. UDP VNI ARP REQ vxlan0 20.0.0.6

3. ff:ff:ff:ff:ff:ff 54:1:C0:A8:64:6 ARP REQ eth0 192.168.100.6
Who has 192.168.100.8 Tell 192.168.100.6 Switch

5. Eth UDP VNI ARP REQ

Ping 20.0.0.8 (on host 6)
1) Host 6 ARP request for 20.0.0.8 on vxlan0 d
2) Host 6 vxlan0 device prepends VXLAN head
3) Host 6 sends ARP request for 192.168.100.
4) Host 8 sends ARP reply for 192.168.100.8 t
5) Host 6 sends encapsulated vxlan0 arp reque
6) Host 8 strips off eth and udp header and for
7) Host 8 vxlan0 device responds to arp reque
8) Host 6 now knows destination MAC address
9) Host 6 now sends ICMP ping request to cor

??? Step 3: How to find out vxlan IF 20.0.0.8 hosted b

29 So

nternet

Host 8 Ping REPLY 20.0.0.6

7. 54:1:14:0:0:8 54:1:14:0:0:6 ARP REPLY
20.0.0.6 is at 54:1:14:0:0:8

vxlan0 20.0.0.8 6. UDP VNI ARP REQ

eth0 192.168.100.8

h 54:1:C0:A8:64:8 54:1:C0:A8:64:6 ARP RPY
4. 192.168.100.8 is at 54:1:C0:A8:64:8

device (broadcast)
der and forwards via UDP to eth0 device
.8 on eth0 device (broadcast)
to host 6 (unicast)
est to host 8 via eth0
rwards to vxlan0 device
est from host 6 vxlan0 device
s of host 8 vxlan0 device
rrect host 8 vxlan mac address

by host 8 (reachable via 192.168.100.8)

oftware Defined Networking using VXLAN, Thomas Richter ([email protected]), LinuxCon 2013

IBM Presentation Template Full Version

VM Attachment and Macvtap Devic

Host B
App A

1 m

macvtap0

NIC

2

Macvtap
■ Combines tun/tap and macvlan devices
■ Modes:

(1)Bridged: destination MAC address
(2)Vepa:Traffic forwarded to external
(3)Private: Same as vepa, but ingres
(4)Passthrough: Only 1 macvtap dev

30 So

ce Options

App

/dev/tapX I/f to User Space (tuntap)

macvtap1 Virtual I/f with new MAC address

34

s lookup on all macvtap devices defined on NIC
l switch
ss traffic blocked
vice allowed per NIC (“exclusive” use)

oftware Defined Networking using VXLAN, Thomas Richter ([email protected]), LinuxCon 2013