Scaling Overlay Virtual Networks

Published on November 2016 | Categories: Documents | Downloads: 29 | Comments: 0 | Views: 251
of 135
Download PDF   Embed   Report

Comments

Content

Scaling Overlay Virtual Networks
Ivan Pepelnjak ([email protected])
Network Architect, ipSpace.net AG
Dimitri Stiliadis ([email protected])
CTO, Nuage Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Past
• CTO of IT and security ventures
• Architect of switches and routers
• Researcher with focus in systems, networking,
and security
Present
• CTO of Nuage Networks
Focus
• Large-scale SDN and cloud environments
• Distributed systems

More @ ipSpace.net/About

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
3 This material
© ipSpace.net
2014
Overlay
Virtual Networks

Past
• Kernel programmer, network OS and web developer
• Sysadmin, database admin, network engineer, CCIE
• Trainer, course developer, curriculum architect
• Team lead, CTO, business owner
Present
• Network architect, consultant, blogger, webinar and book author
• Teaching the art of Scalable Web Application Design
Focus
• Large-scale data centers, clouds and network virtualization
• Scalable application design
• Core IP routing/MPLS, IPv6, VPN
More @ ipSpace.net/About

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
4 This material
© ipSpace.net
2014
Overlay
Virtual Networks









Fully distributed data plane
Scale-out control plane
Availability zones
Hardware gateways
Large-scale microsegmentation
Scaling stateful services
Service chaining

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
5 This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

PHP

Web server

Web server

Web server

App server

App server

Web server

Apache
MySQL
Linux
Cache

Cache
Primary DB

Single VM (LAMP stack)
• Typical SMB deployment
• Simple web hosting

Slave DB

Multi-layer application architecture
• Multiple security zones
• Load balancing and firewalling

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
8 This material
© ipSpace.net
2014
Overlay
Virtual Networks

Outside

Web servers







App servers

DB servers

Multiple logical segments
IP (sometimes MAC) connectivity within a segment
Routing, load balancing and/or firewalling between segments
Baseline firewalling within a segment
Connectivity to the outside world

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
9 This material
© ipSpace.net
2014
Overlay
Virtual Networks

IP packet

MAC unicast
VNI

Encapsulation

VNI

Overlay module
TEP
Kernel IP stack

IP packet

Overlay module
TEP
Kernel IP stack

Hypervisor/Rtr MAC
IP packet

IP transport (underlay) network

• All overlay virtual networking solutions use distributed L2 forwarding
• Scalability is limited by the control plane
(distribution of VM MAC-to-VTEP IP mappings)

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
10This material
© ipSpace.net
2014
Overlay
Virtual Networks

Overlay
Virtual
Network

Outside
Network

Centralized (sometimes VM-based) inter-subnet forwarding doesn’t scale
• Virtual router (L3 agent) becomes a chokepoint
• VM-based forwarding has limited performance
• Avoid this architecture for east-west traffic forwarding
Use architecture with distributed layer-3 forwarding
• Prefer dedicated in-kernel implementation over Linux Kernel TCP/IP stack with
namespaces or VM-based implementations
• Sample products: Juniper Contrail, Microsoft Hyper-V, Nuage VSP, VMware NSX

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
11This material
© ipSpace.net
2014
Overlay
Virtual Networks

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+12This material

ARP: C D
MAC: C  bcast

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
13This
©
2014
Overlay
Virtual Networks
1
ofmaterial
6ipSpace.net

ARP: C D
MAC: C  bcast

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D
• Intercepted by local L3 forwarding module

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
14This
©
2014
Overlay
Virtual Networks
2
ofmaterial
6ipSpace.net

ARP: C D
MAC: C  bcast

ARP: D = MAC-D
MAC: GW  C

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D
• Intercepted by local L3 forwarding module
• Replied from local ARP cache

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
15This
©
2014
Overlay
Virtual Networks
3
ofmaterial
6ipSpace.net

ARP: C D
MAC: C  bcast

ARP: D = MAC-D
MAC: GW  C

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D
• Intercepted by local L3 forwarding module
• Replied from local ARP cache
• Controller is contacted on ARP cache miss

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
16This
©
2014
Overlay
Virtual Networks
4
ofmaterial
6ipSpace.net

ARP: C D
MAC: C  bcast

ARP: D = MAC-D
MAC: GW  C

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D
• Intercepted by local L3 forwarding module
• Replied from local ARP cache
• Controller is contacted on ARP cache miss
• Controller can reply with authoritative information or flood ARP request

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
17This
©
2014
Overlay
Virtual Networks
5
ofmaterial
6ipSpace.net

ARP: C D
MAC: C  bcast

ARP: D = MAC-D
MAC: GW  C

A

VNI: 1

B

C

D

VNI: 2

VNI: 2

Overlay Module

E

F

VNI: 3

Overlay Module

GW

IP (layer-3) transport network

Some overlay virtual networking solutions implement combined L2+L3 forwarding model
• Intra-subnet ARP caching significantly reduces overlay broadcast traffic
Example: ARP request C  D
• Intercepted by local L3 forwarding module
• Replied from local ARP cache
• Controller is contacted on ARP cache miss
• Controller can reply with authoritative information or flood ARP request
Available in VMware NSX for vSphere, Nuage Networks VSP
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
18This
©
2014
Overlay
Virtual Networks
6
ofmaterial
6ipSpace.net

Scaling network services
• Scale-out load balancing is mission impossible
(shared state tied to outside IP address)
• Scale-out firewalls are common
(state tied to a single VM)
• Scale-out NAT is an interesting challenge
Hypervisor

Implement traffic filters with VM NIC firewalls
Outside
• Stateful firewalls or reflexive ACLs
Network
• Reflexive ACLs might be good enough for well-designed
applications
• VM-based solutions severely limit performance
 use in-kernel filters
• Sample solutions: Nuage VSP, VMware NSX, OpenStack/CloudStack on KVM
• ACL-only solutions: Microsoft Hyper-V, VMware vSphere, Cisco Nexus 1000V

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
19This material
© ipSpace.net
2014
Overlay
Virtual Networks

Requirements for scalable data plane
• Distributed L3 forwarding
• Local ARP handling (ARP caching or pure L3 solution)
• Distributed security groups implemented in hypervisors

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
20This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Cloud
Management

Overlay
VTEP
Kernel IP stack

SDN Controller

Overlay
VTEP
Kernel IP stack

IP transport network

Crucial overlay virtual network challenge: VM-MAC-to-VTEP-IP mappings
• Initial implementations used IP multicast and Ethernet-like learning
• Modern solutions use network controllers in combination with orchestration systems
Sample solutions: Cisco Nexus 1000V, Juniper Contrail, Nuage VSP, VMware NSX

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
22This material
© ipSpace.net
2014
Overlay
Virtual Networks

Some overlay networking solutions lack
SDN controller element
• Cloud management platform programs
virtual switches directly
• Hard to integrate with the physical network:
static routes/MAC learning or VM-based
solutions

CMP

Federation
SDN

SDN controller enables inter-cloud
federation
• Reachability data exchanged between
controllers
• Most SDN controllers use BGP for easy integration with existing
hardware

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
23This material
© ipSpace.net
2014
Overlay
Virtual Networks

Overlay
VTEP
Kernel IP stack

Controller

Overlay
VTEP
Kernel IP stack

IP transport network

• Network controller becomes the scalability bottleneck
• Control-plane-only controllers scale much better than controllers participating in
data plane (hint: use CMP to get MAC and IP address information)
• Every controller implementation eventually hits its limits
 scale-out is the only answer

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
24This material
© ipSpace.net
2014
Overlay
Virtual Networks

BGP

Scale-out architecture is the only viable way forward
• Requirement: Synchronization of policy and reachability information between controllers
Typical solution: multi-protocol BGP (MP-BGP)
• L3VPN for IP routing (sometimes using host routes for VM IP addresses)
• EVPN for layer-2 forwarding
• Easy integration with existing hardware gateways
Additional benefits:
• Clean failure domain separation (availability zones)
• Adjustable size of failure domains to meet scalability and convergence requirements
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
25This material
© ipSpace.net
2014
Overlay
Virtual Networks

Terminology:
• VSP: Virtual Services Platform
• CMP: Cloud Management Platform
• VSD: Virtual Services Directory
• VSC: Virtual Services Controller
• VRS: Virtual Routing & Switching
Plane of operation
• VSD: Management/Policy
• VSC: Control plane
• VRS: Data plane

REST
VSD

XMPP

CMP

BGP

BGP

VSC

VRS

VSC

VRS

VRS

VSG/PE

VRS

Scale-out architecture
• Single VSD per CMP
• Multiple VSC per VSD (scale-out within CMP)
• VSC confederation via MP-BGP (scale-out across CMP)
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
26This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Failure Domain: area impacted when a key device or service experiences
problems

Sample failure domains
• VLAN (broadcast storms)
• OSPF area (LSA flooding)
• Controller-based network
(controller failure)
• Cloud instance
(cloud management system failure)

REST
VSD

XMPP

CMP

BGP
VSC

VRS

VSC

VRS

VRS

VRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
28This material
© ipSpace.net
2014
Overlay
Virtual Networks

Regions: cloud instances with separate API endpoints
• Separate instances of cloud management systems

REST
CMP

VSD

XMPP

Availability zone: logical group that provides a
form of physical isolation and redundancy
from other availability zones (OpenStack)
• Common cloud management
• Isolated compute/storage/networking
failure domains
• Each availability zone SHOULD have a
different network services controller

BGP
VSC

VRS

VSC

VRS

VRS

VRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
29This material
© ipSpace.net
2014
Overlay
Virtual Networks

Cloud management platform fails?
• No moves, adds or changes
• Overlay virtual networking topology is frozen
• High-availability clusters cannot recover
SDN controller fails?
• Controllers involved in data plane (MAC learning or
ARP replies)  total failure
• Control-plane controllers  loss of reachability
information
• Controllers without external control plane
 no visibility, no topology change

CMP

CMP

Federation
SDN

Each availability zone SHOULD have an independent
SDN controller

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
30This material
© ipSpace.net
2014
Overlay
Virtual Networks

Underlying infrastructure
• Each availability zone = independent
L3 forwarding domain
VRS

XMPP

Controller/orchestration infrastructure
• Single CMP/VSD per region
• VSD works on policy plane  VSD failure is similar to CMP failure
• VSC per availability zone  VSC failure does not spread across zones
• BGP information exchange through a set of route reflectors 
use BGP security mechanisms to
protect availability zones
REST
• Pair of VSGs per availability zone
CMP
VSD
(when needed)

BGP

BGP

VSC

VSC

VRS

VRS

VSG/PE

VRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
31This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

VMs within an overlay virtual network must interact with the physical world
L2 gateways (VNI-to-VLAN)
• P2V migrations
• Integration with legacy equipment
L3 gateways
• Multiple VNIs routed to a VLAN
• Simple P2V or WAN integration
Network services gateway
• Firewalls and load balancers

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
33This material
© ipSpace.net
2014
Overlay
Virtual Networks

Deployment format
• VM-based
• Hypervisor kernel module
• Bare-metal x86 server
• Hardware VTEP
Design and deployment considerations
• Performance
• Control-plane integration with overlay fabric
• Management plane integration with overlay network controller and
orchestration system
• Integration with existing network infrastructure (example: MPLS/VPN)

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
34This material
© ipSpace.net
2014
Overlay
Virtual Networks

IP packet

VM

IP packet

Appliance MAC

VLAN tag

VXLAN

Next-hop MAC

VNI

UDP

VNI

VXLAN
VTEP
Kernel IP stack

IP multicast

VXLAN
VTEP
Kernel IP stack

MAC multicast

VLAN

IP packet
IP transport network






Outside

Gateway function implemented in a VM with multiple virtual NICs
VM performs traditional bridging/routing/network services functionality
Use any product available in VM format (including Linux instances)
Forwarded traffic goes through a VM  performance usually limited to few Gbps

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
35This material
© ipSpace.net
2014
Overlay
Virtual Networks

Typical gateway deployment scenarios
• Integrate overlay networks with outside world
 maximum performance = WAN link speed
• Integrate overlay networks with legacy hardware
 maximum performance = legacy hardware network I/O performance
Software gateway performance
• Few Gbps for VM-based solutions
• ~10Gbps for kernel-based and bare-metal gateways

Hardware gateways offer the performance needed in large-scale deployments
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
36This material
© ipSpace.net
2014
Overlay
Virtual Networks

Hardware Gateway needs the following information
• Mapping between VXLAN VNI and external VLANs
• VM-MAC-to-VTEP-IP mappings
• VXLAN flooding information (IP MC address or VTEP list)
Solutions
• Do-it-yourself
• OVSDB (VMware NSX, Nuage VSP)
• EVPN (Nuage VSP, Juniper Contrail)

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
37This material
© ipSpace.net
2014
Overlay
Virtual Networks

OVSDB
• Lightweight JSON-RPC-based database query/update protocol
• OVSDB database table schema defines the actual data

Hardware VTEP schema
• Physical switch + ports
• Logical switch + router
• Local and remote MAC mappings
SDN controller uses OVSDB to
• Configure VXLAN-to-VLAN mappings
• Push MAC mappings to VTEP
• Receive physical MAC addresses
from VTEP

OVSDB

MPLS/VPN integration through VLANs (Inter-AS Option A)
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
38This material
© ipSpace.net
2014
Overlay
Virtual Networks

• Network virtualization controller and
hardware gateway use EVPN and L3VPN
to exchange forwarding data
• EVPN provides MAC-to-VTEP mappings
• L3VPN provides integrates overlay
virtual networks with MPLS/VPN
• Gateway provisioning uses
a different protocol (ex: NETCONF)

EVPN
L3VPN

EVPN forwarding information
• VTEP flood list (Inclusive Multicast Ethernet Tag route)
• MAC-to-VTEP mapping (MAC/IP Address Advertisement route)
• Propagation of IP addresses enables proxy ARP functionality

MPLS/VPN integration through MP-BGP (same domain or inter-AS Option B/C)
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
39This material
© ipSpace.net
2014
Overlay
Virtual Networks

MPLS/VPN
GW

VSC

PE

Nuage VRS

Underlay IP transport network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+40This material

MPLS/VPN

MP-BGP
GW

VSC

PE

Nuage VRS

Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
41This
©
2014
Overlay
Virtual Networks
1
ofmaterial
7ipSpace.net

OpenFlow
GW

MPLS/VPN

MP-BGP

VSC

PE

Nuage VRS

Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
42This
©
2014
Overlay
Virtual Networks
2
ofmaterial
7ipSpace.net

IP: A  S
MAC: A  GW
OpenFlow
GW

MPLS/VPN

MP-BGP

VSC

PE

Nuage VRS

Underlay IP transport network

• PE-router sends VPNv4 or EVPN

• VM sends IP packet to server (and GW MAC)

update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
43This
©
2014
Overlay
Virtual Networks
3
ofmaterial
7ipSpace.net

IP: A  S
MAC: A  GW
OpenFlow
GW

MPLS/VPN

MP-BGP

VSC

PE

Nuage VRS

Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

• VM sends IP packet to server (and GW MAC)
• IP router in VRS performs L3 lookup

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
44This
©
2014
Overlay
Virtual Networks
4
ofmaterial
7ipSpace.net

OpenFlow
IP: A  S
GW
Nuage VRS

MPLS label

MPLS/VPN

MP-BGP

VSC

PE

GRE header
IP to PE

Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

• VM sends IP packet to server (and GW MAC)
• IP router in VRS performs L3 lookup
• IP packet is encapsulated in MPLS-GRE-IP or
VXLAN-UDP-IP envelope

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
45This
©
2014
Overlay
Virtual Networks
5
ofmaterial
7ipSpace.net

OpenFlow
GW

MPLS/VPN

MP-BGP

VSC

PE

Nuage VRS
IP to PE VTEP
Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

• VM sends IP packet to server (and GW MAC)
• IP router in VRS performs L3 lookup
• IP packet is encapsulated in MPLS-GRE-IP or
VXLAN-UDP-IP envelope
• PE router receives MPLS/VPN or VXLAN
packet

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
46This
©
2014
Overlay
Virtual Networks
6
ofmaterial
7ipSpace.net

OpenFlow
GW
Nuage VRS

MPLS/VPN

MP-BGP

VSC

PE
IP/MPLS to S

Underlay IP transport network

• PE-router sends VPNv4 or EVPN
update to Nuage VSC
• VSC installs forwarding entries with
BGP next hop + label in VRS

• VM sends IP packet to server (and GW MAC)
• IP router in VRS performs L3 lookup
• IP packet is encapsulated in MPLS-GRE-IP or
VXLAN-UDP-IP envelope
• PE router receives MPLS/VPN or VXLAN
packet
• PE router forwards VPN IP packet

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
47This
©
2014
Overlay
Virtual Networks
7
ofmaterial
7ipSpace.net

Deployment format
• Low bandwidth  VM
• High bandwidth  hardware VTEP
Integration requirements
• Physical VLANs  OVSDB or EVPN
• MPLS/VPN WAN  EVPN + L3VPN

Choose an SDN controller that supports all the options you need
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
48This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Security Groups Concepts
• Replace subnet-level firewalls (or ACLs) with per-VM firewalls/ACLs
• Increased intra-subnet security due to microsegmentation
• No chokepoint, no traffic tromboning
• No subnets  no addressing limitations
Implementations
• CloudStack (on Linux-based hypervisors)
• OpenStack (Neutron plugin extension)
• VMware vCD/vCAC with vShield Edge or
VMware NSX

Outside
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
50This material
© ipSpace.net
2014
Overlay
Virtual Networks

High-level view
• Assign VMs to groups
• Specify filtering rules between groups

From

To

Any

Web

80

Any

Web

443

Typical implementations
• Packet filter (OVS or Linux iptables)
• Each group exploded into a list of IP addresses
• ACL = Cartesian product of source-destination
IP addresses

Web

App

9000

App

DB

3306

Mgmt

All-VM

Port

22

Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
51This material
© ipSpace.net
2014
Overlay
Virtual Networks

From

To

Any

Web

Any

From

To

80

Any

W1

80

Web

443

Any

W2

80

Web

App

9000

Any

W3

80

App

DB

3306

Any

W1

443

Mgmt

All-VM

22

Any

W2

443

Any

W3

443

W1

A1

9000

W1

A2

9000

W2

A1

9000

W2

A2

9000

W3

A1

9000

W3

A2

9000

W1

W2

Port

W3

A1

D1

Outside

A2

D2

Port



is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
52This material
© ipSpace.net
2014
Overlay
Virtual Networks

Security group ACL = Cartesian product of IP
addresses
• Long ACLs (performance usually degrades
linearly with the ACL length)
• Whole ACL deployed on all VM NICs
 even further performance degradation
• Any change in security group membership
(VM adds or removals) propagates to all
hypevisors running tenant’s VMs

SDN
Hypervisor
Outside
Network

From

To

Port

Any

Web

80

Any

Web

443

Web

App

9000

App

DB

3306

Mgmt

All-VM

22

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
53This material
© ipSpace.net
2014
Overlay
Virtual Networks

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+54This material

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
55This
©
2014
Overlay
Virtual Networks
1
ofmaterial
6ipSpace.net

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
56This
©
2014
Overlay
Virtual Networks
2
ofmaterial
6ipSpace.net

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
57This
©
2014
Overlay
Virtual Networks
3
ofmaterial
6ipSpace.net

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
58This
©
2014
Overlay
Virtual Networks
4
ofmaterial
6ipSpace.net

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
59This
©
2014
Overlay
Virtual Networks
5
ofmaterial
6ipSpace.net

Security group membership = BGP community
• Remote VM security group attached to IP or MAC route
• Local VM security group attached to VM port

VSD

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)
• ACL is not changed

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
60This
©
2014
Overlay
Virtual Networks
6
ofmaterial
6ipSpace.net

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+61This material

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
62This
©
2014
Overlay
Virtual Networks
1
ofmaterial
5ipSpace.net

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)
• VSC originates new EVPN and IPVPN route
(security group = BGP community)

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
63This
©
2014
Overlay
Virtual Networks
2
ofmaterial
5ipSpace.net

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)
• VSC originates new EVPN and IPVPN route
(security group = BGP community)
• VSC sends BGP update to its
BGP peers

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
64This
©
2014
Overlay
Virtual Networks
3
ofmaterial
5ipSpace.net

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)
• VSC originates new EVPN and IPVPN route
(security group = BGP community)
• VSC sends BGP update to its
BGP peers
• Remote VSC updates forwarding
VRS
entries in remote VRS

VSD

VSC

VSC

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
65This
©
2014
Overlay
Virtual Networks
4
ofmaterial
5ipSpace.net

Typical sequence of events
• New VM is started on a hypervisor
• VRS notifies VSC
• VSC notifies VSD
• VSD assigns VM into a security group and
replies to VSC
• VSC updates MAC-to-VTEP and IP-to-VTEP
forwarding entries (incl. security group)
• VSC originates new EVPN and IPVPN route
(security group = BGP community)
• VSC sends BGP update to its
BGP peers
• Remote VSC updates forwarding
VRS
entries in remote VRS
• ACL is not changed

VSD

VSC

VSC

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
66This
©
2014
Overlay
Virtual Networks
5
ofmaterial
5ipSpace.net

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+67This material

VM sends an IP packet

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
68This
©
2014
Overlay
Virtual Networks
1
ofmaterial
5ipSpace.net

VM sends an IP packet
Ingress ACL check on ingress VRS
• From security group = VM NIC group
• To security group = BGP community
VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
69This
©
2014
Overlay
Virtual Networks
2
ofmaterial
5ipSpace.net

VM sends an IP packet
Ingress ACL check on ingress VRS
• From security group = VM NIC group
• To security group = BGP community
Encapsulated VM frame is sent across the
transport network

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
70This
©
2014
Overlay
Virtual Networks
3
ofmaterial
5ipSpace.net

VM sends an IP packet
Ingress ACL check on ingress VRS
• From security group = VM NIC group
• To security group = BGP community
Encapsulated VM frame is sent across the
transport network
Egress ACL check on egress VRS
• From security group = BGP community
• To security group = VM NIC group

VSD

VSC

VSC

VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
71This
©
2014
Overlay
Virtual Networks
4
ofmaterial
5ipSpace.net

VM sends an IP packet
Ingress ACL check on ingress VRS
• From security group = VM NIC group
• To security group = BGP community
Encapsulated VM frame is sent across the
transport network
Egress ACL check on egress VRS
• From security group = BGP community
• To security group = VM NIC group

VSD

VSC

VSC

Packet is delivered to target VM
VRS

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
72This
©
2014
Overlay
Virtual Networks
5
ofmaterial
5ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+73This material

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
74This
©
2014
Overlay
Virtual Networks
1
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• Packet delivered to VM

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
75This
©
2014
Overlay
Virtual Networks
2
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• Packet delivered to VM
VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
76This
©
2014
Overlay
Virtual Networks
3
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• IP packet delivered to remote host
• Packet delivered to VM
VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
77This
©
2014
Overlay
Virtual Networks
4
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• IP packet delivered to remote host

Remote host to VM:
• IP packet received by PE-router
• Packet delivered to VM

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
78This
©
2014
Overlay
Virtual Networks
5
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• IP packet delivered to remote host

Remote host to VM:
• IP packet received by PE-router
• IP packet delivered to VRS
• Packet delivered to VM

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
79This
©
2014
Overlay
Virtual Networks
6
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• IP packet delivered to remote host

Remote host to VM:
• IP packet received by PE-router
• IP packet delivered to VRS
• Egress ACL on VRS

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
80This
©
2014
Overlay
Virtual Networks
7
ofmaterial
8ipSpace.net

Security groups (in BGP communities) can extend across MPLS/VPN backbone
• Automatic ingress/egress filters on VM NICs
• Requires trust (or strict filters) between cloud and MPLS/VPN networks
VM to remote host:
• VM sends a packet
• Ingress ACL on VRS
• IP packet sent from VRS to PE-router
• IP packet delivered to remote host

Remote host to VM:
• IP packet received by PE-router
• IP packet delivered to VRS
• Egress ACL on VRS
• Packet delivered to VM

VSC

MPLS
backbone

VRS
Transport Network

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
81This
©
2014
Overlay
Virtual Networks
8
ofmaterial
8ipSpace.net

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Shared state

Scale-out NAT is hard problem
• No guarantee of symmetrical paths
(Best case: rehashing after topology change)
• Shared state tied to outside IP address
• State must be distributed and synchronized across all NAT cluster
members
Maybe we’re solving the wrong problem
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
83This material
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Floating IP address

NAT

• Virtual machines with public IP addresses (Floating IP address)
 static stateless NAT
• Access to outside servers
 dynamic stateful NAPT, outside source address is irrelevant

Equivalent to Amazon VPC behavior
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
85This material
© ipSpace.net
2014
Overlay
Virtual Networks

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP
Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+86This material

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
87This
©
2014
Overlay
Virtual Networks
1
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
88This
©
2014
Overlay
Virtual Networks
2
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
89This
©
2014
Overlay
Virtual Networks
3
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)
• Hypervisor translates destination IP address to VM IP address

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
90This
©
2014
Overlay
Virtual Networks
4
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)
• Hypervisor translates destination IP address to VM IP address
Inside-to-outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
91This
©
2014
Overlay
Virtual Networks
5
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)
• Hypervisor translates destination IP address to VM IP address
Inside-to-outside
• VM sends packet to a destination unreachable in tenant vDRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
92This
©
2014
Overlay
Virtual Networks
6
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)
• Hypervisor translates destination IP address to VM IP address
Inside-to-outside
• VM sends packet to a destination unreachable in tenant vDRS
• Per-VM default route pushes the packet through NAT rule into public vDRS

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
93This
©
2014
Overlay
Virtual Networks
7
ofmaterial
8ipSpace.net

Setup
• Floating IP from public vDRS is
allocated to a tenant VM
• 1:1 NAT rule is created on the
hypervisor

Tenant vDRS (VRF)
F-IP

Internal communication
• Destination IP address is within tenant vDRS
• NAT rule is not invoked

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Packet sent to IP address in public vDRS (received by hypervisor)
• Hypervisor translates destination IP address to VM IP address
Inside-to-outside
• VM sends packet to a destination unreachable in tenant vDRS
• Per-VM default route pushes the packet through NAT rule into public vDRS

NAT rule is stateless and active on a single hypervisor
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
94This
©
2014
Overlay
Virtual Networks
8
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor

Tenant vDRS (VRF)
H-IP
Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
© ipSpace.net
2014
Overlay
Virtual Networks
+95This material

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS

Tenant vDRS (VRF)
H-IP
Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
96This
©
2014
Overlay
Virtual Networks
1
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS

Tenant vDRS (VRF)
H-IP
Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
97This
©
2014
Overlay
Virtual Networks
2
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
98This
©
2014
Overlay
Virtual Networks
3
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor
• Packet is delivered to the outside server

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
99This
©
2014
Overlay
Virtual Networks
4
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor
• Packet is delivered to the outside server

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Return packet is sent to IP address in public vDRS (received by hypervisor)

This
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
100
©
2014
Overlay
Virtual Networks
5
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor
• Packet is delivered to the outside server

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Return packet is sent to IP address in public vDRS (received by hypervisor)
• Hypervisor uses PNAT entry to translate destination IP address to VM IP address

This
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
101
©
2014
Overlay
Virtual Networks
6
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor
• Packet is delivered to the outside server

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Return packet is sent to IP address in public vDRS (received by hypervisor)
• Hypervisor uses PNAT entry to translate destination IP address to VM IP address
• Translated packet is delivered to target VM

This
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
102
©
2014
Overlay
Virtual Networks
7
ofmaterial
8ipSpace.net

Setup
• IP from public vDRS (H-IP) is
allocated to each hypervisor
H-IP
Inside-to-outside
• VM sends packet to a destination
unreachable in tenant vDRS
• Default route pushes the packet
through NAT rule into public vDRS
• Stateful NAT entry is created in the hypervisor
• Packet is delivered to the outside server

Tenant vDRS (VRF)

Public
vDRS
(VRF)
Transport
Network
VSG/PE
Outside

Outside-to-inside
• Return packet is sent to IP address in public vDRS (received by hypervisor)
• Hypervisor uses PNAT entry to translate destination IP address to VM IP address
• Translated packet is delivered to target VM

The goal is connectivity, not specific NAT outside address
This
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
103
©
2014
Overlay
Virtual Networks
8
ofmaterial
8ipSpace.net

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars






Insert physical appliances between virtual network endpoints
Insert L4-7 and security services within a subnet
Create multi-tier applications without routing overhead
Combine multiple services in Network Function Virtualization
deployments

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
105
© ipSpace.net
2014
Overlay
Virtual Networks

A
S
B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
106
© ipSpace.net
2014
Overlay
Virtual Networks
+

A

IP-A  IP-S

MAC-A  MAC-S

S
B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

1 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
107
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

S
B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

2 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
108
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

S
IP-B  IP-S

MAC-B  MAC-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

3 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
109
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

S
IP-B  IP-S

MAC-B  MAC-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

4 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
110
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

S
IP-B  IP-S

MAC-B  MAC-S

IP-B  IP-S

MAC-B  MAC-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

5 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
111
© ipSpace.net
2014
Overlay
Virtual Networks

MAC-A  MAC-S

A

IP-A  IP-S

S
IP-B  IP-S

MAC-B  MAC-S

IP-B  IP-S

MAC-B  MAC-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

6 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
112
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

S
IP-B  IP-S

MAC-B  MAC-S

IP-B  IP-S

MAC-B  MAC-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

7 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
113
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

S
MAC-B  MAC-S IP-B  IP-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

8 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
114
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

S
MAC-B  MAC-S IP-B  IP-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

9 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
115
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

S
MAC-B  MAC-S IP-B  IP-S

MAC-B  MAC-S IP-B  IP-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers

10 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
116
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-S

IP-A  IP-S

MAC-A  MAC-S

IP-A  IP-S

S
MAC-B  MAC-S IP-B  IP-S

MAC-B  MAC-S IP-B  IP-S

B

Layer-2 frames redirected to a transparent (bump-in-wire) appliance
• Based on MAC (potentially IP) headers
Typical implementation
• VLAN chaining
• Hard to implement for individual endpoints
• Impossible to implement for individual applications
• Fantastic potential for forwarding loops

11 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
117
© ipSpace.net
2014
Overlay
Virtual Networks

A
S
B

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
118
© ipSpace.net
2014
Overlay
Virtual Networks
+

A

IP-A  IP-S

MAC-A  MAC-G

S
B

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

1 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
119
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-G

IP-A  IP-S

MAC-G  MAC-S

S
B

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

2 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
120
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

MAC-A  MAC-G

IP-A  IP-S

MAC-G  MAC-S

S
IP-B  IP-S

MAC-B  MAC-G

B

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

3 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
121
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

B

MAC-B  MAC-G

IP-A  IP-S

MAC-G  MAC-S

 MAC-F

IP-B  IP-S

MAC-A  MAC-G

S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

4 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
122
© ipSpace.net
2014
Overlay
Virtual Networks

A

IP-A  IP-S

B

MAC-B  MAC-G

IP-A  IP-S

 MAC-F

IP-B  IP-S

MAC-A  MAC-G

MAC-G  MAC-S

S
IP-B  IP-S

MAC-F  MAC-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

5 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
123
© ipSpace.net
2014
Overlay
Virtual Networks

MAC-G  MAC-S IP-A  IP-S

A

B

MAC-B  MAC-G

 MAC-F

IP-B  IP-S

S
IP-B  IP-S

MAC-F  MAC-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

6 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
124
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-G IP-A  IP-S

B

MAC-B  MAC-G

 MAC-F

IP-B  IP-S

MAC-G  MAC-S IP-A  IP-S

S
IP-B  IP-S

MAC-F  MAC-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

7 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
125
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-G IP-A  IP-S

MAC-G  MAC-S IP-A  IP-S

S
MAC-G  MAC-S IP-B  IP-S

B

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

8 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
126
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-G  MAC-S IP-A  IP-S

 MAC-F

B

MAC-A  MAC-G IP-A  IP-S

S
MAC-G  MAC-S IP-B  IP-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

9 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
127
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-G IP-A  IP-S

B

IP-B  IP-S

 MAC-F

MAC-B  MAC-F

MAC-G  MAC-S IP-A  IP-S

S
MAC-G  MAC-S IP-B  IP-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite

10 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
128
© ipSpace.net
2014
Overlay
Virtual Networks

A

MAC-A  MAC-G IP-A  IP-S

B

IP-B  IP-S

 MAC-F

MAC-B  MAC-F

MAC-G  MAC-S IP-A  IP-S

S
MAC-G  MAC-S IP-B  IP-S

Layer-3 frames redirected to a transparent or inter-subnet appliance
• Based on IP headers
• Might require MAC header rewrite
Typical implementation
• Policy-based routing (PBR)
• MAC rewrite is automatic
• Hard to implement for appliances not close to the forwarding path

11 of 11

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
129
© ipSpace.net
2014
Overlay
Virtual Networks

• Services and redirection (chaining) rules are defined in VSD Architect
• VSD downloads redirection rules to VSC
• VSC instantiates PBR entries on
virtual port (VM) activation
• Traffic redirection uses the same
scalability mechanisms as
security groups
• Multiple forwarding domains are used
to further scale the implementation

VSD

VSC

VSC

VRS

VRS
Transport Network

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
130
© ipSpace.net
2014
Overlay
Virtual Networks

• Appliances (physical or virtual) are identified by virtual port tags
• A dedicated VNI (VXLAN segment) is allocated to each appliance port
• Appliance reachability information (ESI, VNI, transport next hop) is
propagated in EVPN updates
• Information from EVPN update is used as PBR next hop

VSC

VSC
MP-BGP

VRS

VRS
Transport Network

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
131
© ipSpace.net
2014
Overlay
Virtual Networks








GARP

Appliances (physical or virtual) are identified by virtual port tags
A dedicated VNI (VXLAN segment) is allocated to each appliance port
L2VPN is create between appliance
Active appliance IP address is detected by monitoring GARP packets
A host route is created for each appliance IP address
L3VPN host route (prefix, VNI, transport
MP-BGP
next hop) toward appliance port is
propagated across MP-BGP
VSC
VSC
routing domain
• Information from L3VPN route is
used as PBR next hop
VRS

VRS

VRS

Transport Network

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
132
© ipSpace.net
2014
Overlay
Virtual Networks

This material is copyrighted and licensed for the sole use by Dimitar Stojanovski ([email protected] [164.143.240.34]). More information at http://www.ipSpace.net/Webinars

Architectural elements:
• Distributed forwarding plane (L2 and L3)
• Control plane with scale-out architecture
• Distributed L4 services (security, NAT)
• Scalable security mechanisms
Additional considerations:
• High-performance gateways
• Control- and management-plane integration with external networks

This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
134
© ipSpace.net
2014
Overlay
Virtual Networks

• Define the services
• Define the virtual infrastructure requirements
 Connectivity (L2 and/or L3)
 Security
 Performance
 Integration with legacy infrastructure
 Integration with WAN networks
• Select the orchestration system
• Select the hypervisor platform
• Select an overlay virtual networking solution that will support the services
you want to offer
 Easy integration with the orchestration system
 Scalable implementation of network services
 Scalable integration with external networks
This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
135
© ipSpace.net
2014
Overlay
Virtual Networks

Questions?

Send them to [email protected] or @ioshints
This material
is copyrighted
and licensed for the sole use by Dimitar Scaling
Stojanovski
([email protected]
[164.143.240.34]). More information at http://www.ipSpace.net/Webinars
136
© ipSpace.net
2014
Overlay
Virtual Networks

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close