OpenFlow - Sándor Laki

Download Report

Transcript OpenFlow - Sándor Laki

Software Defined Network
and
Network Virtualization
Sándor Laki
(Slides by Yeh-Ching Chung)
Introduction
Motivation
Concept
Open Flow
Virtual Switch
SOFTWARE DEFINED NETWORK
We have lost our way
Routing, management, mobility management,
access control, VPNs, …
App
App
App
Operating
System
Specialized Packet
Forwarding Hardware
Million of lines
of source code
5400 RFCs
Barrier to entry
500M gates
10Gbytes RAM
Bloated
Power Hungry
Many complex functions baked into the infrastructure
OSPF, BGP, multicast, differentiated services,
Traffic Engineering, NAT, firewalls, MPLS, redundant layers, …
An industry with a “mainframe-mentality”
Reality
App
App
App
App
Operating
System
App
App
Operating System
Specialized Packet
Forwarding Hardware
Specialized Packet
Forwarding Hardware
• Lack of competition means glacial innovation
• Closed architecture means blurry, closed interfaces
• Vertically integrated, complex, closed, proprietary
• Not suitable for experimental ideas
• Not good for network owners & users
• Not good for researchers
Glacial process of innovation made worse
by captive standards process
Deployment
Idea
Standardize
Wait 10 years
•
•
•
•
Driven by vendors
Consumers largely locked out
Lowest common denominator features
Glacial innovation
Introduction
Motivation
Concept
Open Flow
Virtual Switch
SOFTWARE DEFINED NETWORK
Trend
App
App
App
Windows
Windows
Windows
(OS)
(OS)
(OS)
Linux
Linux
Linux
App
App
App
Mac
Mac
Mac
OS
OS
OS
Virtualization layer
x86
(Computer)
Computer Industry
Controller11
NOX
Controller
(Network OS)
Controller
Controller
Network
OS
22
Virtualization or “Slicing”
OpenFlow
Network Industry
The “Software-defined Network”
App
App
App
Network Operating System
Ap
p
Ap
p
Ap
p
Operating
System
Ap
p
Specialized Packet
Forwarding Hardware
Ap
p
Ap
p
Ap
p
Ap
p
Operating
System
Ap
p
Specialized Packet
Forwarding Hardware
Operating
System
Ap
p
Specialized Packet
Forwarding Hardware
Ap
p
Ap
p
Operating
System
Ap
p
Ap
p
Ap
p
Operating
System
Specialized Packet
Forwarding Hardware
Specialized Packet
Forwarding Hardware
The “Software-defined Network”
2. At least one good operating system
Extensible, possibly open-source
3. Well-defined open API
App
App
App
Network Operating System
1. Open interface to hardware
Simple Packet
Forwarding
Hardware
Simple Packet
Forwarding
Hardware
Simple Packet
Forwarding
Hardware
Simple Packet
Forwarding
Hardware
Simple Packet
Forwarding
Hardware
Isolated “slices”
App
App
Network
Operating
System 1
Many operating systems, or
Many versions
App
App
Network
Operating
System 2
App
App
App
Network
Operating
System 3
App
Network
Operating
System 4
Open interface to hardware
Virtualization or “Slicing” Layer
Open interface to hardware
Simple Packet
Forwarding Hardware
Simple Packet
Forwarding Hardware
Simple Packet
Forwarding Hardware
Simple Packet
Forwarding Hardware
Simple Packet
Forwarding Hardware
Consequences
More innovation in network services
– Owners, operators, 3rd party developers,
researchers can improve the network
– E.g. energy management, data center
management, policy routing, access control,
denial of service, mobility
Lower barrier to entry for competition
– Healthier market place, new players
Introduction
Motivation
Concept
Open Flow
Virtual Switch
SOFTWARE DEFINED NETWORK
Traditional network node: Router
• Router can be partitioned into control and data plane
– Management plane/ configuration
– Control plane / Decision: OSPF (Open Shortest Path First)
– Data plane / Forwarding
Adjacent Router
Routing
Control plane
OSPF
Switching
Data plane
Router
Management/Policy plane
Configuration / CLI / GUI
Static routes
Control plane
OSPF
Neighbor
table
Data plane
Link state
database
Adjacent Router
Control plane
OSPF
IP routing
table
Forwarding table
Data plane
Traditional network node: Switch
• Typical Networking Software
– Management plane
– Control Plane – The brain/decision maker
– Data Plane – Packet forwarder
SDN Concept
• Separate Control plane and Data plane entities
– Network intelligence and state are logically centralized
– The underlying network infrastructure is abstracted from the
applications
• Execute or run Control plane software on general
purpose hardware
– Decouple from specific networking hardware
– Use commodity servers
• Have programmable data planes
– Maintain, control and program data plane state from a central
entity
• An architecture to control not just a networking device
but an entire network
Control Program
Control program operates on view of network
– Input: global network view (graph/database)
– Output: configuration of each network device
Control program is not a distributed system
– Abstraction hides details of distributed state
Software-Defined Network with key
Abstractions in the Control Plane
Network
Virtualization
Well-defined API
Routing
Traffic
Engineering
Other
Applications
Network Operating System
Separation of Data
and Control Plane
Forwarding
Forwarding
Forwarding
Forwarding
Network Map
Abstraction
Forwarding Abstraction
Purpose: Abstract away forwarding hardware
Flexible
– Behavior specified by control plane
– Built from basic set of forwarding primitives
Minimal
– Streamlined for speed and low-power
– Control program not vendor-specific
OpenFlow is an example of such an abstraction
OpenFlow Basics
Control Program A
Control Program B
Network OS
OpenFlow Protocol
Ethernet
Switch
Control
Path
OpenFlow
Data Path (Hardware)
OpenFlow Basics
Control Program A
Control Program B
Network OS
“If header = p, send to port 4”
Packet
Forwarding
Packet
Forwarding
“If header = q, overwrite header with r,
add header s, and send to ports 5,6”
“If header = ?, send to me”
Flow
Table(s)
Packet
Forwarding
Plumbing Primitives
<Match, Action>
Match arbitrary bits in headers:
Header
Data
Match: 1000x01xx0101001x
–
–
Match on any header, or new header
Allows any flow granularity
Action
–
–
–
Forward to port(s), drop, send to controller
Overwrite header with mask, push or pop
Forward at specific bit-rate
21
General Forwarding Abstraction
Small set of primitives
“Forwarding instruction set”
Protocol independent
Backward compatible
Switches, routers, WiFi APs,
basestations, TDM/WDM
Introduction
Motivation
Concept
Open Flow
Virtual Switch
SOFTWARE DEFINED NETWORK
What is OpenFlow
• OpenFlow is similar to an x86 instruction set for the
network
• Provide open interface to “black box” networking node
– (ie. Routers, L2/L3 switch) to enable visibility and openness in
network
• Separation of control plane and data plane.
– The datapath of an OpenFlow Switch consists of a Flow Table,
and an action associated with each flow entry
– The control path consists of a controller which programs the
flow entry in the flow table
• OpenFlow is based on an Ethernet switch, with an
internal flow-table, and a standardized interface to add
and remove flow entries
OpenFlow Consortium
http://OpenFlowSwitch.org
• Goal
– Evangelize OpenFlow to vendors
– Free membership for all researchers
– Whitepaper, OpenFlow Switch Specification,
Reference Designs
– Licensing: Free for research and commercial
use
OpenFlow building blocks
oftrace
oflops
Monitoring/
debugging tools
openseer
Stanford Provided
ENVI (GUI)
NOX
LAVI
Beacon
FlowVisor
Console
n-Casting
Trema
Applications
ONIX
Controller
Maestro
Slicing
Software
FlowVisor
Stanford Provided
Commercial Switches
HP, NEC, Pronto,
Juniper.. and many
more
Expedient
Software
Ref. Switch
NetFPGA
Broadcom
Ref. Switch
OpenWRT
PCEngine
WiFi AP
Open vSwitch
OpenFlow
Switches
26
Components of OpenFlow Network
• Controller
– OpenFlow protocol messages
– Controlled channel
– Processing
• Pipeline Processing
• Packet Matching
• Instructions & Action Set
• OpenFlow switch
– Secure Channel (SC)
– Flow Table
• Flow entry
OpenFlow Controllers
Name
Lang
Platform(s)
License
Original
Author
Notes
OpenFlow
Reference
C
Linux
OpenFlow
License
Stanford/Nici
ra
not designed for extensibility
NOX
Pytho
n, C++
Linux
GPL
Nicira
actively developed
Beacon
Java
Win, Mac,
Linux,
Android
GPL (core),
David
FOSS Licenses Erickson
for your code (Stanford)
Maestro
Java
Win, Mac,
Linux
LGPL
Zheng Cai
(Rice)
Trema
Ruby,
C
Linux
GPL
NEC
includes emulator, regression test
framework
OpenDaylight
Java
Linux
?
OpenDaylight
Community
Linux
Effort is supported by large vendors
runtime modular, web UI framework,
regression test framework
28
Secure Channel (SC)
• SC is the interface that connects each OpenFlow switch to
controller
• A controller configures and manages the switch via this interface.
– Receives events from the switch
– Send packets out the switch
• SC establishes and terminates the connection between
OpneFlow Switch and the controller using the procedures
– Connection Setup
– Connection Interrupt
• The SC connection is a TLS connection. Switch and controller
mutually authenticate by exchanging certificates signed by a
site-specific private key.
Flow Table
• Flow table in switches, routers, and chipsets
Flow 1.
Rule
(exact & wildcard)
Action
Statistics
Flow 2.
Rule
(exact & wildcard)
Action
Statistics
Flow 3.
Rule
(exact & wildcard)
Action
Statistics
Flow N.
Rule
(exact & wildcard)
Default Action
Statistics
Flow Entry
• A flow entry consists of
– Match fields
• Match against packets
– Action
• Modify the action set or pipeline processing
– Stats
Match
Fields
• Update the matching packets
In Port
Src
MAC
Dst
MAC
Eth
Type
Vlan Id
Layer 2
1.
2.
3.
4.
Forward packet to port(s)
Encapsulate and forward to controller
Drop packet
Send to normal processing pipeline
IP Tos
IP
Proto
IP Src
Layer 3
Action
IP Dst
Stats
TCP Src
Port
TCP Dst
Port
Layer 4
1. Packet
2. Byte counters
Examples
Switching
Switch MAC
Port src
*
MAC Eth
dst
type
00:1f:.. *
*
VLAN IP
ID
Src
IP
Dst
IP
Prot
TCP
TCP
Action
sport dport
*
*
*
*
IP
Dst
IP
Prot
TCP
TCP
Action
sport dport
*
*
port6
Flow Switching
Switch MAC
Port src
MAC Eth
dst
type
port3 00:20.. 00:1f.. 0800
VLAN IP
ID
Src
vlan1 1.2.3.4 5.6.7.8
4
17264 80
port6
Firewall
Switch MAC
Port src
*
*
MAC Eth
dst
type
*
*
VLAN IP
ID
Src
IP
Dst
IP
Prot
TCP
TCP
Action
sport dport
*
*
*
*
*
22
drop
32
Examples
Routing
Switch MAC
Port src
*
*
MAC Eth
dst
type
*
*
VLAN IP
ID
Src
IP
Dst
*
5.6.7.8 *
*
IP
Dst
TCP
TCP
Action
sport dport
*
IP
Prot
TCP
TCP
Action
sport dport
*
port6
VLAN Switching
Switch MAC
Port src
*
*
MAC Eth
dst
type
00:1f.. *
VLAN IP
ID
Src
vlan1 *
*
IP
Prot
*
*
*
port6,
port7,
port9
33
OpenFlow Usage
Controller
Peter’s code
OpenFlow
Rule Switch
Action
PC
Statistics
OpenFlow
Protocol
OpenFlow
Action
Switch
Rule
OpenFlowSwitch.org
Statistics
OpenFlow
Action
Switch
Rule
Peter
Statistics
Usage examples
• Peter’s code:
– Static “VLANs”
– His own new routing protocol: unicast, multicast, multipath, loadbalancing
– Network access control
– Home network manager
– Mobility manager
– Energy manager
– Packet processor (in controller)
– IPvPeter
– Network measurement and visualization
– …
Separate VLANs for Production and
Research Traffic
Controller
Research VLANs
Flow Table
Production VLANs
Normal L2/L3 Processing
Dynamic Flow Aggregation on an OpenFlow
Network
• Scope
– Different Networks want different flow granularity (ISP,
Backbone,…)
– Switch resources are limited (flow entries, memory)
– Network management is hard
– Current Solutions : MPLS, IP aggregation
Dynamic Flow Aggregation on an OpenFlow
Network
• How do OpenFlow Help?
– Dynamically define flow granularity by wildcarding
arbitrary header fields
– Granularity is on the switch flow entries, no packet rewrite
or encapsulation
– Create meaningful bundles and manage them using your
own software (reroute, monitor)
Virtualizing OpenFlow
• Network operators “Delegate” control of
subsets of network hardware and/or traffic to
other network operators or users
• Multiple controllers can talk to the same set of
switches
• Imagine a hypervisor for network equipments
• Allow experiments to be run on the network
in isolation of each other and production
traffic
Switch Based Virtualization
Exists for NEC, HP switches but not flexible enough
Research VLAN 2
Flow Table
Controller
Research VLAN 1
Flow Table
Controller
Production VLANs
Normal L2/L3 Processing
40
FlowVisor
• A network hypervisor developed by Stanford
• A software proxy between the forwarding and
control planes of network devices
FlowVisor-based Virtualization
Heidi’s
Controller
Aaron’s
Controller
Craig’s
Controller
Topology
discovery is
per slice
OpenFlow
Protocol
OpenFlow
Switch
OpenFlow FlowVisor
& Policy Control
OpenFlow
Protocol
OpenFlow
Switch
OpenFlow
Switch
42
FlowVisor-based Virtualization
Separation not only
by VLANs, but any
L1-L4 pattern
Broadcast
Multicast
OpenFlow
Protocol
dl_dst=FFFFFFFFFFFF
OpenFlow
Switch
http
Load-balancer
tp_src=80, or
tp_dst=80
OpenFlow
FlowVisor & Policy Control
OpenFlow
Protocol
OpenFlow
Switch
OpenFlow
Switch
43
FlowVisor Slicing
• Slices are defined using a slice definition policy
– The policy language specifies the slice’s resource
limits, flowspace, and controller’s location in
terms of IP and TCP port-pair
– FlowVisor enforces transparency and isolation
between slices by inspecting, rewriting, and
policing OpenFlow messages as they pass
FlowVisor Resource Limits
• FV assigns hardware resources to “Slices”
– Topology
• Network Device or Openflow Instance (DPID)
• Physical Ports
– Bandwidth
• Each slice can be assigned a per port queue with a fraction
of the total bandwidth
– CPU
• Employs Course Rate Limiting techniques to keep new flow
events from one slice from overrunning the CPU
– Forwarding Tables
• Each slice has a finite quota of forwarding rules per device
Slicing
FlowVisor FlowSpace
• FlowSpace is defined by a collection of packet
headers and assigned to “Slices”
– Source/Destination MAC address
– VLAN ID
– Ethertype
– IP protocol
– Source/Destination IP address
– ToS/DSCP
– Source/Destination port number
FlowSpace: Maps Packets to Slices
FlowVisor Slicing Policy
• FV intercepts OF messages from devices
– FV only sends control plane messages to the Slice
controller if the source device is in the Slice
topology.
– Rewrites OF feature negotiation messages so the
slice controller only sees the ports in it’s slice
– Port up/down messages are pruned and only
forwarded to affected slices
FlowVisor Slicing Policy
• FV intercepts OF messages from controllers
– Rewrites flow insertion, deletion & modification rules
so they don’t violate the slice definition
• Flow definition – ex. Limit Control to HTTP traffic only
• Actions – ex. Limit forwarding to only ports in the slice
– Expand Flow rules into multiple rules to fit policy
• Flow definition – ex. If there is a policy for John’s HTTP traffic
and another for Uwe’s HTTP traffic, FV would expand a single
rule intended to control all HTTP traffic into 2 rules.
• Actions – ex. Rule action is send out all ports. FV will create
one rule for each port in the slice.
– Returns “action is invalid” error if trying to control a
port outside of the slice
FlowVisor Message Handling
Alice
Controller
Bob
Controller
Cathy
Controller
OpenFlow
Policy Check:
Is this rule
allowed?
Policy Check:
Who controls
this packet?
FlowVisor
OpenFlow
Full Line Rate
Forwarding
Packet
Packet
OpenFlow
Firmware
Data Path
Rule
Exception
Introduction
Motivation
Concept
Open Flow
Virtual Switch
SOFTWARE DEFINED NETWORK
INTRODUCTION
• Due to the cloud computing service, the number
of virtual switches begins to expand dramatically
– Management complexity, security issues and even
performance degradation
• Software/hardware based virtual switches as well
as integration of open-source hypervisor with
virtual switch technology is exhibited
53
Software-Based Virtual Switch
• The hypervisors implement
vSwitch
• Each VM has at least one
virtual network interface
cards (vNICs) and shared
physical network interface
cards (pNICs) on the physical
host through vSwitch
• Administrators don’t have
effective solution to
separate packets from
different VM users
• For VMs reside in the same
physical machine, their
traffic visibility is a big issue
54
Issues of Traditional vSwitch
• The traditional vSwitches lack of advanced
networking features such as VLAN, port mirror,
port channel, etc.
• Some hypervisor vSwitch vendors provide
technologies to fix the above problems
– OpenvSwitch may be superior in quality for the
reasons
55
Open vSwitch
• A software-based solution
– Resolve the problems of network separation and
traffic visibility, so the cloud users can be assigned
VMs with elastic and secure network configurations
• Flexible Controller in User-Space
• Fast Datapath in Kernel
Server
Open vSwitch Controller
Open vSwitch Datapath
Open vSwitch Concepts
• Multiple ports to physical switches
– A port may have one or more interfaces
• Bonding allows more than once interface per port
• Packets are forwarded by flow
• Visibility
– NetFlow
– sFlow
– Mirroring (SPAN/RSPAN/ERSPAN)
• IEEE 802.1Q Support
– Enable virtual LAN function
– By attaching VLAN ID to Linux virtual interfaces, each
user will have its own LAN environment separated
from other users
Open vSwitch Concepts
• Fine-grained ACLs and QoS policies
– L2-‐L4 matching
– Actions to forward, drop, modify, and queue
– HTB and HFSC queuing disciplines
• Centralized control through OpenFlow
• Works on Linux-based hypervisors:
– Xen
– XenServer
– KVM
– VirtualBox
Open vSwitch Contributors(Partial)
Packets are Managed as Flows
• A flow may be identied by any combination of
– Input port
– VLAN ID (802.1Q)
– Ethernet Source MAC address
– Ethernet Destination MAC address
– IP Source MAC address
– IP Destination MAC address
– TCP/UDP/... Source Port
– TCP/UDP/... Destination Port
Packets are Managed as Flows
• The 1st packet of a flow is sent to the controller
• The controller programs the datapath's actions
for a flow
– Usually one, but may be a list
– Actions include:
•
•
•
•
Forward to a port or ports
mirror
Encapsulate and forward to controller
Drop
• And returns the packet to the datapath
• Subsequent packets are handled directly by the
datapath
Migration
• KVM and Xen provide Live Migration
• With bridging, IP address migration must
occur with in the same L2 network
• Open vSwitch avoids this problem using GRE
tunnels
Hardware-Based Virtual Switch
• Why hardware-based?
– Software virtual switches consume CPU and
memory usage
– Possible inconsistence of network and server
configurations may cause errors and is very hard
to troubleshooting and maintenance
• Hardware-based virtual switch solution emerges
for better resource utilization and configuration
consistency
63
Virtual Ethernet Port Aggregator
• A standard led by HP, Extreme, IBM, Brocade,
Juniper, etc.
• An emerging technology as part of IEEE
802.1Qbg Edge Virtual Bridge (EVB) standard
• The main goal of VEPA is to allow traffic of
VMs to exit and re-enter the same server
physical port to enable switching among VMs
64
Virtual Ethernet Port Aggregator
• VEPA software update is required
for host servers in order to force
packets to be transmitted to
external switches
• An external VEPA enabled switch is
required for communications
between VMs in the same server
• VEPA supports “hairpin” mode
which allows traffic to “hairpin”
back out the same port it just
received it from--- requires
firmware update to existing
switches
65
Pros. and Cons. for VEPA
• Pros
– Minor software/firmware update, network
configuration maintained by external switches
• Cons
– VEPA still consumes server resources in order to
perform forwarding table lookup
66
References
•
•
•
•
•
•
•
•
•
•
•
•
"OpenFlow: Enabling Innovation in Campus Networks“ N. McKeown, T. Andershnan, G. Parulkar,
L. Peterson, J. Rexford, S. Shenker, and J. Turneron, H. Balakris ACM Computer Communication
Review, Vol. 38, Issue 2, pp. 69-74 April 2008
OpenFlow Switch Specication V 1.1.0.
Richard Wang, Dana Butnariu, and Jennifer Rexford OpenFlow-based server load balancing gone
wild, Workshop on Hot Topics in Management of Internet, Cloud, and Enterprise 66 IP Infusion
Proprietary and Confidential, released under Customer NDA , Roadmap items subject to change
without notice © 2011 IP Infusion Inc. gone wild, Workshop on Hot Topics in Management of
Internet, Cloud, and Enterprise Networks and Services (Hot-ICE), Boston, MA, March 2011.
Saurav Das, Guru Parulkar, Preeti Singh, Daniel Getachew, Lyndon Ong, Nick McKeown, Packet
and Circuit Network Convergence with OpenFlow, Optical Fiber Conference (OFC/NFOEC'10), San
Diego, March 2010
Nikhil Handigol, Srini Seetharaman, Mario Flajslik, Nick McKeown, Ramesh Johari, Plug-n-Serve:
Load-Balancing Web Traffic using OpenFlow, ACM SIGCOMM Demo, Aug 2009.
NOX: Towards an Operating System for Networks
https://sites.google.com/site/routeflow/home
http://www.openflow.org/
http://www.opennetsummit.org/
https://www.opennetworking.org/
http://conferences.sigcomm.org/sigcomm/2010/papers/sigcomm/p195.pdf
http://searchnetworking.techtarget.com/
References
• Network Virtualization with Cloud Virtual Switch
• S. Horman, “An Introduction to Open vSwitch,”
LinuxCon Japan, Yokohama, Jun. 2, 2011.
• J. Pettit, J. Gross “Open vSwitch Overview,” Linux
Collaboration Summit, San Francisco, Apr. 7, 2011.
• J. Pettit, “Open vSwitch: A Whirlwind Tour,” Mar. 3,
2011.
• Access Layer Network Virtualization: VN-Tag and VEPA
• OpenFlow Tutorial
Network Virtualization
Network Design Rules
• Hierarchical approach
– Traffic is aggregated
hierarchically from an access
layer into a layer of
distribution switches and
finally onto the network core.
– A hierarchical approach to
network design has proven to
deliver the best results in
terms of optimizing
scalability, improving
manageability, and
maximizing network
availability.
Network Virtualization
• What is network virtualization ?
71
Network Virtualization
• What is network virtualization ?
– In computing, network virtualization is the process of
combining hardware and software network resources and
network functionality into a single, software-based
administrative entity, a virtual network.
• Two categories :
– External network virtualization
• Combine many networks, or parts of networks, into a virtual unit.
– Internal network virtualization
• Provide network-like functionality to the software containers on a
single system.
Network Virtualization
• Desirable properties of network virtualization :
– Scalability
• Easy to extend resources in need
• Administrator can dynamically create or delete virtual network
connection
– Resilience
• Recover from the failures
• Virtual network will automatically redirect packets by redundant
links
– Security
• Increased path isolation and user segmentation
• Virtual network should work with firewall software
– Availability
• Access network resource anytime
73
Network Virtualization
• External network virtualization in different layers :
– Layer 1
• Seldom virtualization implement in this physical data transmission
layer.
– Layer 2
• Use some tags in MAC address packet to provide virtualization.
• Example, VLAN.
– Layer 3
• Use some tunnel techniques to form a virtual network.
• Example, VPN.
– Layer 4 or higher
• Build up some overlay network for some application.
• Example, P2P.
Network Virtualization
• Internal network virtualization in different layers :
– Layer 1
• Hypervisor usually do not need to emulate the physical layer.
– Layer 2
• Implement virtual L2 network devices, such as switch, in hypervisor.
• Example, Linux TAP driver + Linux bridge.
– Layer 3
• Implement virtual L3 network devices, such as router, in hypervisor.
• Example, Linux TUN driver + Linux bridge + iptables.
– Layer 4 or higher
• Layer 4 or higher layers virtualization is usually implemented in guest
OS.
• Applications should make their own choice.
Introduction
External network virtualization
Internal network virtualization
NETWORK VIRTUALIZATION
Network Virtualization
• Two virtualization
components :
– Device virtualization
• Virtualize physical devices in the
network
– Data path virtualization
• Virtualize communication path
between network access points
Switch
Data Path
Router
77
Network Virtualization
• Device virtualization
– Layer 2 solution
• Divide physical switch
into multiple logical
switches.
 Layer 3 solution 3
• VRF technique
( Virtual Routing and Forwarding )
• Emulate isolated routing tables
within one physical router.
78
Network Virtualization
• Data path virtualization
– Hop-to-hop case
• Consider the
virtualization applied on a
single hop data-path.
– Hop-to-cloud case
• Consider the
virtualization tunnels
allow multi-hop datapath.
79
Network Virtualization
• Protocol approach
– Protocols usually use for data-path virtualization.
– Three implementations
• 802.1Q – implement hop to hop data-path
virtualization
• MPLS ( Multiprotocol Label Switch ) – implement
router and switch layer virtualization
• GRE (Generic Routing Encapsulation ) – implement
virtualization among wide variety of networks with
tunneling technique.
80
Network Virtualization
• 802.1Q
– Standard by IEEE 802.1
– Not encapsulate the
original frame
– Add a 32-bit field between
MAC address and
EtherTypes field
• ETYPE(2B): Protocol
identifier
• Dot1Q Tag(2B): VLAN
number, Priority code
CE: Customer Edge router
PE: Provider Edge router
81
Network Virtualization
• Example of 802.1Q
VN 1
Source
destination
Physical Network
VN 2
Source
destination
82
Network Virtualization
• MPLS ( Multiprotocol Label Switch )
– Also classified as layer 2.5 virtualization
– Add one or more labels into package
– Need Label Switch Router(LSR) to read MPLS header
83
Packet Traversing a
Label-Switched Path
Network Virtualization
• Example of MPLS
5
4
VN 1
2
7
9
8
LSR
LER
CE
Physical Network
LER
LSR
CE
LER
CE
5
4
7
2
VN 2
9
85
Network Virtualization
• GRE ( Generic Routing Encapsulation )
– GRE is a tunnel protocol developed by CISCO
– Encapsulate a wide variety of network layer protocols inside virtual
point-to-point links over an Internet Protocol internetwork
– Stateless property
• This means end-point doesn't keep information about the state
Built Tunnel
86
Introduction
External network virtualization
Internal network virtualization
NETWORK VIRTUALIZATION
Internal Network Virtualization
• Internal network virtualization
– A single system is configured with containers, such as the
Xen domain, combined with hypervisor control programs
or pseudo-interfaces such as the VNIC, to create a
“network in a box”.
– This solution improves overall efficiency of a single system
by isolating applications to separate containers and/or
pseudo interfaces.
– Virtual machine and virtual switch :
• The VMs are connected logically to each other so that they can
send data to and receive data from each other.
• Each virtual network is serviced by a single virtual switch.
• A virtual network can be connected to a physical network by
associating one or more network adapters (uplink adapters) with
the virtual switch.
Internal Network Virtualization
• Properties of virtual switch
– A virtual switch works much like a physical Ethernet switch.
– It detects which VMs are logically connected to each of its
virtual ports and uses that information to forward traffic to
the correct virtual machines.
• Typical virtual network configuration
– Communication network
• Connect VMs on different hosts
– Storage network
• Connect VMs to remote storage system
– Management network
• Individual links for system administration
Internal Network Virtualization
Network virtualization example form VMware
KVM Approach
• In KVM system
– KVM focuses on CPU and memory virtualization, so IO
virtualization framework is completed by QEMU.
– In QEMU, network interface of virtual machines
connect to host by TUN/TAP driver and Linux bridge.
• Virtual machines connect to host by a virtual network
adapter, which is implemented by TUN/TAP driver.
• Virtual adapters will connect to Linux bridges, which play the
role of virtual switch.
KVM Approach
• TUN/TAP driver
– TUN and TAP are virtual network kernel drivers :
• TAP (as in network tap) simulates an Ethernet device and operates
with layer 2 packets such as Ethernet frames.
• TUN (as in network TUNnel) simulates a network layer device and
operates with layer 3 packets such as IP.
– Data flow of TUN/TAP driver
• Packets sent by an operating system via a TUN/TAP device are
delivered to a user-space program that attaches itself to the
device.
• A user-space program may pass packets into a TUN/TAP device.
TUN/TAP device delivers (or "injects") these packets to the
operating system network stack thus emulating their reception
from an external source.
KVM Approach
KVM Approach
• Linux bridge
– Bridging is a forwarding technique used in packet-switched computer
networks.
– Unlike routing, bridging makes no assumptions about where in a
network a particular address is located.
– Bridging depends on flooding and examination of source addresses in
received packet headers to locate unknown devices.
– Bridging connects multiple network
segments at the data link layer
(Layer 2) of the OSI model.
KVM Approach
TAP/TUN driver + Linux Bridge
Xen Approach
• In Xen system
– Since implemented by para-virtualization, guest OS
loads modified network interface drivers.
– Modified network interface drivers, which act as TAP
in KVM approach, communicate with virtual switches
in Dom0.
– Virtual switch in Xen can be
implemented by Linux bridge
or work with other approaches.
Network Virtualization Summary
• Virtualization in layers
– Usually in Layer 2 and Layer 3
• External network virtualization
– Layer 2
• 802.1q
– Layer 3
• MPLS, GRE
• Internal network virtualization
– Traditional approach
• TAP/TUN + Linux bridge
– New technique
• Virtual switch
Reference
• Books :
– Kumar Reddy & Victor Moreno, Network Virtualization, Cisco Press 2006
• Web resources :
– Linux Bridge http://www.ibm.com/developerworks/cn/linux/l-tuntap/index.html
– Xen networking http://wiki.xensource.com/xenwiki/XenNetworking
– VMware Virtual Networking Concepts
http://www.vmware.com/files/pdf/virtual_networking_concepts.pdf
– TUN/TAP wiki http://en.wikipedia.org/wiki/TUN/TAP
– Network Virtualization wiki http://en.wikipedia.org/wiki/Network_virtualization
• Papers :
– A. Menon, A. Cox, and W. Zwaenepoel. Optimizing Network Virtualization in Xen. Proc.
USENIX Annual Technical Conference (USENIX 2006), pages 15–28, 2006.