PE-based IPv6 multicast transition for mesh

problem

Yong Cui @ CERNET

2

Why we need multicast transition? From CERNET

Existing multicast applications are in IPv4

Native IPv6 CERNET2 is expected to support IPv4 multicast applications

IPv4 multicast traffic is low

3

What’s multicast transition problem

IPv6 backbone

I Pv4 Customer

I Pv4 Customer I Pv4 Customer

CECE

PE

PE

CE

PE

Source

Receiver Receiver

How to setup a multicast tree in IPv6 backbone where edge of tree should be PE?

4

Auto-discovery members in mesh framework Auto-discovery of PEs (N PEs totally)

Mesh framework or 4over6 based on MP-BGP

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

P

MP-B

GP

MP-BGP

MP

-B

GP

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CE

MP-BGP

5

Potential solution Construct multicast tree in IPv6 backbon

eDynamic mapped IPv6 tree (M trees)

Dynamically mapping IPv4 mgroup to IPv6 mgroup

Single static tree by PIM-SM IPv6 RP in IPv6 backbone

PE-based static tree by PIM-SM (N trees totally)

One tree rooted by each PE SPT rooted by each PE to all other PE

6

IPv4/v6 group mapping

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

P

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CE

RPv4

(S,G)v4 <->(S’,G’)v6(S,G)v4 <->(S’,G’)v6

(S,G)v4 <->(S’,G’)v6 ,S’ is IPv6 address of the PE of S G’ is ff18:ssss:ssss:gggg:gggg::x

s and g are the (S,G) P maintains M IPv6 group states if there are M IPv4

groups

7

Construct a single tree in IPv6 backbone PE4 members table

PE IPv6 addr

PE1 IFv AFy

PE2 IFv AFy

be joined join

OK OK

OK OK

PE3 IFv AFy OK OK

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

RPv6

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CEPE1 members table

PE IPv6 addr

PE2 IFv AFy

PE3 IFv AFy

be joined join

OK OK

OK OK

PE4 IFv AFy OK OK

•A single IPv6 group address is configure on RPv6 and PEs•Egress PE initiates a single join procedure to RPv6•SPT rooted by RPv6, IPv6 tree optimization is a problem•The number of IPv6 multicast groups is 1

8

Construct one tree rooted by each PE in IPv6 backbonePE4 members table

PE IPv6 addr

PE1 IFv AFy

PE2 IFv AFy

be joined join

OK OK

OK OK

PE3 IFv AFy OK OK

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

P

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CEPE1 members table

PE IPv6 addr

PE2 IFv AFy

PE3 IFv AFy

be joined join

OK OK

OK OK

PE4 IFv AFy OK OK

•Each ingress PE has its own IPv6 group address •ingress PE sends its IPv6 group address to all other PEs by Softwire extensions•Egress PE initiates a join procedure to each ingress PE (SPT rooted by each PE)•The number of IPv6 multicast groups is N

9

IPv4 PIM-SM Sender/receiver Registration over IPv6 backbone

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

P

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CE

RPv4

S tree

S register

Source

10

Multicast Forwarding

CE

Ipv4 customer

CEPE1 PE4

PE2

IPv6 backbone

P

Ipv4 customer

Ipv4 customer

PE3CE

Ipv4 customer

CE

RPv4

Source

11

Construct multicast tree in IPv6 backbone Dynamic mapped IPv6 tree (M trees)

Dynamically mapping IPv4 mgroup to IPv6 mgroup Scalability problem in IPv6 core

Single static tree rooted by IPv6 RP IPv6 tree optimization is a problem

PE-based static tree by PIM-SM (N trees totally) One tree rooted by each PE SPT rooted by each PE to all other PE Scalability: Keep less states in the backbone Efficiency: PE may receive some packets they do not requi

re Future work

Comments & co-authors are welcome To be a WG document

Conclusion and future work