BIER. Bit Indexed Explicit Replica0on. MBONED, IETF 92 Greg Shepherd

Transcription

1 BIER Bit Indexed Explicit Replica0on MBONED, IETF 92 Greg Shepherd

2 The BIER Epiphany Only encode the end- receivers in the packet header. Not the intermediate nodes. Assign end- receivers a Bit Posi0on from a Bit String. The smallest iden0fier possible. Encode the Bit String in the packet header. Using some sort of encapsula0on. Create a Bit Forwarding Table on all BIER nodes to allow mul0cast packet forwarding using the Bit String in the packet. Derived from the RIB, SPF based. We call it, Bit Indexed Explicit Replica0on (BIER).

3 IETF The BIER idea was presented in a BOF at the IETF in Hawaii. November 2014 First IETF WG Mee0ng, March 2015 Vendors collabora0ng (presen0ng) with us; Ericsson Alcatel- Lucent Juniper Huawei Received very good trac0on, driving this to a new Working Group

4 BIER IETF BIER Charter: h]ps://datatracker.ie_.org/wg/ bier/charter/ IETF 92 Hack- a- thon Best Hardcore Hack Winner: Markus Stenberg BIER in OpenWRT for Homnet

5 IETF dra`s dra`- shepherd- bier- problem- statement dra`- wijnands- bier- architecture dra`- wijnands- mpls- bier- encapsula0on dra`- kumar- bier- use- cases dra`- rosen- l3vpn- mvpn- bier dra`- psenak- ospf- bier- extensions dra`- przygienda- bier- isis- ranges

6 Solu0on Overview

7 Basic Idea BIER A/32 B/32 LSA 1 - A/32 LSA 2 B/32 BIER Domain LSA 5 E/32 LSA 3 C/32 LSA 4 D/32 E/ BitString 1 C/32 D/32 1. Assign a unique Bit Posi0on from a BitString to each BFER in the BIER domain. 2. Each BFER floods their Bit Posi0on to BFR- prefix mapping using the IGP (OSPF, ISIS)

8 Basic Idea BIER BIER Domain BitMask 0011 A 0100 B 1000 C 1. Assign a unique Bit Posi0on from a mask to each edge router in the BIER domain. 2. Each edge router floods their bit- posi0on- to- ID mapping with a new LSA OSPF or ISIS 3. All BFR s use unicast RIB to calculate a best path for each BFR- prefix 4. Bit Posi0ons are OR d together to form a Bit Mask per BFR- nbr 5. Packets are forwarded and replicated hop- by- hop using the Bit Forwarding Table..

9 Bit Index Forwarding Table BM BM BM 0111 B 0011 C 0100 E 0001 D 0010 F D 0001 BM- ER A B C BM 0011 C B E 0100 BM- ER F 0010 BM- ER D, F and E adver0se their Bit posi0ons in the IGP (flooded). A, B and C know the mapping between the Bit and RID, Based on shortest path route to RID, the Bit Mask Forwarding Table is created.

10 Forwarding Packets Overlay session BM BM BM 0001 AND B A & C & D &0001 AND AND 0100 E 0010 F 0001 B 0001 C D C B E 0100 F 0010

11 Forwarding Packets Overlay session AND B & C & D &0001 AND AND 0100 E & F 0001 A 0101 B 0001 C D 0001 AND C B E 0100 F 0010

12 Forwarding Packets Overlay session AND B & C & D &0001 AND AND 0100 E & F & A B 0011 C D 0001 E F 0010 AND 0011 C B

13 Forwarding Packets As you can see from the previous slides, the result from the bitwise AND (&) between the Bit Mask in the packet and the Forwarding table is copied in the packet for each neighbor. This is the key mechanism to prevent duplica0on. Look at the next slide to see what happens if the bits are not reset If the previous bits would not have been reset, E would forward the packet to C and vice versa.

14 Forwarding Packets Overlay session AND B & A 0100 AND C 0100 E B &0011 &0100 AND D 0010 F C D E 0100 F 0010 AND 0011 C B

15 How many Bits and Where? The number of mul0cast egress routers that can be addressed is depending on the number of Bits that can be included in the BitString The BitString length is depending on the encapsula0on type and router pla_orm. We iden0fied 5 different encoding op0ons, most a]rac0ve below; 1. MPLS, below the bo]om label and before IP header. 2. IPv6, extensions header.

16 MPLS encapsula0on Top Label Bo]om Label BIER Header Payload Normal label from Pla_orm Label Space The last label in the MPLS header. The BIER header encoded between MPLS and IP header Start of the encapsulated packet Top Label Bo]om Label BIER Header VPN Label Payload MPLS Label stack from top to bo]om Between MPLS and IP IPv4/IPv6/L Proto Len Entropy BitString (first 32 bits) ~ ~ ~ ~ BitString (last 32 bits) Reserved BFIR-id Proto = 1: MPLS packet with downstream-assigned label at top of stack.

17 MVPN over BIER

18 MVPN over BIER BIER replaces PIM, mldp, RSVP- TE or IR in the core. BIER represents a full mesh (P2MP) connec0vity between all the PE s in the network. There is no need to explicitly signal any MDT s (or PMSI s). With MVPN there are many profiles, This is partly due to the tradeoff between State and Flooding. Different C- mul0cast signaling op0ons. MVPN over BIER, there is one profile. BGP for C- mul0cast signaling. No need for Data- MDTs.

19 MVPN over BIER (S,G) PIM (*,G):0:0001 (*,G):0:0010 (*,G):0:0001 RR PIM (*,G) (S1,G) PIM A 0100 C 0001 PIM (*,G) BIER (*,G):0:0001 (S2,G) PIM B D PIM (*,G) The BGP control plane defined for MVPN can be re- used. PIM (S,G)/(*,G) can be translated into BGP updates. Requirement, we depend on Leaf AD routes for explicit tracking! Big difference, there is no Tree per VPN!!! The BIER packets needs to carry Source ID and upstream VPN context label

20 Sets and Areas

21 BIER Sets To increase the scale we group the egress routers in Sets. Each Bit Posi0on is unique in the context of a give Set The packet carries the Set ID. Set BM I I J 1:0111 2:0111 Note, we create different forwarding entries for each Set I G H A B C D E 1:0001 1:0010 1:0100 2:0001 2:0010 Set 1 Set 2 Note, Bit Posi0ons 1,2,3 appear in both Sets, jet do not overlap due to Sets. F 2:0100

22 BIER Sets There is no topological restric0on which set an egress belongs to But it may be more efficient if it follows the topology Set BM I G A B 1:0001 Set 1 1: I J 1:0111 2:0111 I Set 2 C 2:0001 Note, we create different forwarding entries for each Set H Set 1 D E F 1:0100 2:0010 Set 2 2:0100

23 BIER Sets If a mul0cast flow has mul0ple receivers in different Sets, the packet needs to be replicated mul0ple 0mes by the ingress router, for each set once. Is that a problem? We don t think so The Set iden0fier is part of the packet. Can be implemented as MPLS label.

24 BIER Area A bit Mask only needs to be unique in its own area. ABR s translate Bit Masks between area s. Requires a IP lookup and state on the ABRs. This is very similar for Segmented Inter- AS MVPN. BM BM BM BM 0:10 ABR 0:01 A 0:01 B 0:10 ABR Area 1 Area 2 A ABR B {0:01} {0:10} {0:10} {0:01}

25 Interdomain BIER A bit Mask only needs to be unique in its own area. Requires a IP lookup and state on peering routers. P2P Peering link could use PIM or IGMP/MLD to join outside BIER domain P2MP Peering exchanges could run BIER, peering routers then opperate as ABRs, but no PIM signaling between domains. BM BM BM BM 0:10 B 0:01 A 0:01 B 0:10 B SP1 SP2 A B B A {0:01} {0:10} {0:10} {0:01}

26 BIER Forwarding

27 BIER Forwarding, neighbor based. BIER Forwarding Table Packet IN 1 1 A & 1 B & C 1 1 D & = = Packet OUT Packet OUT When a packet is received with a BitString, it needs to be matched against each Neighbor in the BFT to determine which neighbor(s) to replicate the packet to. This model works well for systems that have mul0- cores and allow parallel processing. Each neighbor can be processed independently in parallel. Its works less well for serialized processing as it requires a neighbor walk for each packet.

28 BIER Forwarding, Bit Indexed BFT Indexed BFT Neighbors A 1 B 1 C 1 1 D D A B C D We translate the BFT neighbor table in to a table sor0ng on Bit Posi0on (and not by neighbor). We walk the Bit Mask in the packet and Index into the FIB table.

29 BIER Forwarding, Bit Indexed Packet IN Walk Bit Mask & Copy & BIER FIB Label Forwarding Table D A B C = Packet OUT D We walk the Bits in the packet, as soon as we hit a 1, we copy the packet, index into the FIB table with the posi0on of the Bit. The Bit Mask entry is reverse & with the Bit Mask in the packet. This resets the Bits that where processed.

30 BIER Forwarding, Bit Indexed Packet IN Walk Bit Mask & Copy 1 BIER FIB Label Forwarding Table & D = Copy & A B C Packet OUT = Packet OUT D We walk the Bits in the packet, as soon as we hit a 1, we copy the packet, index into the FIB table with the posi0on of the Bit. The Bit Mask entry is reverse & with the Bit Mask in the packet. This resets the Bits that where processed.

31 BIER Forwarding, Bit Indexed Walking the bits in a Bit String takes less clock cycles compared to walking a list of neighbors. For that reason its faster to walk the Bit String and index into the neighbor table. The table is a NxN bit matrix, where N is the Bit String length. Bits that where already processed are reset so we don t processes them if they appears later in a Bit String. This way we avoid mul0ple copies being forwarded. Kudo s to John Besnk!

32 BIER Forwarding, Bit Indexed Walking the Bit String in the packet is basically a repea0ng of Find First Set Bit opera0on. Could be op0mized to record the last posi0on. h]p://en.wikipedia.org/wiki/find_first_set h]p://graphics.stanford.edu/~seander/bithacks.html Checkout Bruijn sequence (Count the consecu0ve zero bits) Is supported in compilers ffs and can be done in HW fairly easily.

33 Conclusions

34 Advantages Packets forwarded via BIER follow the unicast path towards the receiver, inheri0ng unicast features like FRR and LFA. There is no per mul0cast flow state in the network. Mul0cast convergence is as fast as unicast, there is no mul0cast state to re- converge, signal, etc. Nice plugin for SDN, its only the ingress and egress that need to exchange Sender and Receiver informa0on. The core network provides a many- 2- many connec0vely between all BIER routers by default following the IGP. No Mul0cast control protocol in the network.

35 Disadvantages The Bit String length has an upper bound and may not cover all deployment scenarios. Using sets to increase the number of egress routers may cause the ingress to replicate the packet mul0ple 0mes. Using area s requires the ABR to have state. Exis0ng low- end pla_orms are less flexible to adopt BIER.

36 Ques0ons?

37