Transcript 4_IGP_Multicast_Lucy

IGP Multicast Architecture
draft-yong-rtgwg-igp-mutlicast-arch-01
Lucy Yong, Weiguo Hao, Donald Eastlake
Andrew Qu, Jon Hudson, Uma Chunduri
February 2015 NVO3 Interim Meeting
Motivation
• Trend is to decouple network IP space from service IP space in
Data Center environment
– Benefits: networking agility and programmability to applications that are
in IP and non-IP space
– Service IP space is known as overlay net., network IP as underlay net.
• If network IP space is decoupled from service IP space,
– Underlay IP network itself no longer needs manual configuration
– Automatic formation of an IP network fabric can be done (i.e. underlay IP)
– IP network fabric can be simplified by reducing protocols
• IP network fabric needs support unicast and multicast transport
– IGP protocol already supports unicast
– IGP protocol does not yet support multicast, simple extension will do
February 2015
NVO3 Interim Meeting
2
IGP Multicast Architecture
• Support IP multicast packets delivery in an IGP
network
• Use single IGP protocol for unicast and
multicast packets delivery
• No change on the IGP protocol for unicast
packet routing and forwarding
• Use of existing hardware, i.e., data plane
• Require simple IGP protocol extension for IP
multicast packet delivery
February 2015
NVO3 Interim Meeting
3
IGP Multicast Architecture
• A set of edge routers and transit routers:
– They are IGP routers and support IGP multicast
– Multicast source(s) and receiver(s) connect to at least one edge router
– IGP multicast supports multi-tenant environment (IP or non-IP)
• tenants can use the same multicast address space independently
R3
R1
R4
R7
R1-R6 are edge routers
R7-R9 are transit routers
R8
R9
R2
R5
R6
February 2015
Example: A IGP Network with Multicast Capability
NVO3 Interim Meeting
4
IP Multicast Packet Delivery over a Tree
• An algorithm selects a default tree root node or operator can
configure the tree root(s)
• Routers in IGP compute rooted distrib. tree(s) based on LSDB
• Each multicast group associates to at least one distribution tree
• A pruned tree is done based on edge router membership on a
multicast group (optimal transport)
R3
R3
R1
R1
R4
R4
R7
R7
R8
R8
R9
R2
R5
R5
R6
R6
R7 Rooted Distribution Tree (Green)
February 2015
R9
R2
Pruned tree for (*,G) w/[R1,R4,R6] (Green)
NVO3 Interim Meeting
5
Tenant Multicast Packet Delivery
• Tenant (overlay) multicast packets can be L3 or L2 (BUM) packets
• Tenant multicast packets are encap. at ingress NVE prior to
forwarding over IGP Multicast Network
• The mapping b/w tenant multicast family and a underlay (*,G) can
be configured or done by the policy at NVEs
• NVE and IGP Edge Router may be co-located or separated
• Tenant multicast receivers send and/or reply an IGMP/MLD for
joining/leaving a multicast family to local NVE
NVE
Server
February 2015
Edge
Router
IGP Multicast
Network
NVE
Server
NVO3 Interim Meeting
Edge
Router
IGP Multicast
Network
6
Summary
• NVO3 is a primary use case for IGP Multicast
• IGP Mcast development will be in MCAST WG
– IGP Multicast Architecture
• Support single area and multiple areas
• Support IS-IS and OSPF networks
• Support both IPv4 and IPv6 networks
– IGP Multicast Solution (Algorithms and Process
Procedures)
– IS-IS and OSPF extensions for IGP Multicast
February 2015
NVO3 Interim Meeting
7