tb2377 michelet trill vs spb_final


IEEE or IETF

TRILL or SPB

Philippe Michelet, Director of Global Product Management, Data Center Core Switching June 2012

© Copyright 2012 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. 3

Roadmap and is subject to change without notice.

Disclaimer

This document contains forward looking statements regarding future operations, product development, product capabilities and availability dates. This information is subject to substantial uncertainties and is subject to change at any time without prior notification. Statements contained in this document concerning these matters only reflect Hewlett Packard's predictions and / or expectations as of the date of this document and actual results and future plans of Hewlett-Packard may differ significantly as a result of, among other things, changes in product strategy resulting from technological, internal corporate, market and other changes. This is not a commitment to deliver any material, code or functionality and should not be relied upon in making purchasing decisions.


Agenda

1. Problem statement

2. Solution A, IEEE: PBB, PBB-TE, SPB

3. Solution B, IETF: TRILL

4. HPN’s position / roadmap

5. Conclusion


Consider Evolutions Since STP STP like protocols – how high is your blood pressure?

Limited CAPEX Links in standby mode?

Limited OPEX Teams spending weeks to

design the network?

Network = critical resource

Waiting tens of seconds

between failovers ?

Highly virtualized 1000 servers, 50VMs

Does it scale?

Multi-tenancy Can you isolate traffic

between “tenants”?


Legacy STP verses Modern Architectures

Blocked links/idle infrastructure / no multi-pathing

Complex to engineer (STP/RSTP/MSTP)

Slow re-convergence after failover (best case ~1s – typically 3, worse case 45s)

Edge

Aggregation

Core Optimal paths

Actual path


Requirements for a Modern Layer 2 Network


Modern Architecture Requirements

• Arbitrary topologies • All links active, all the time • Multi-pathing/load splitting among multiple paths • Unicast, Multicast and Broadcast support • Compatible with IEEE 802.1 Ethernet networks using STP • Very minimal configuration required • Uncompromised stability


IEEE: SPB (aka 802.1aq)


IEEE Layer 2 Protocol History

Payload

EtherType

SA

DA

IEEE 802.1 Payload

EtherType

C-TAG

EtherType

SA

DA

VLANs IEEE 802.1Q

Payload

EtherType

C-TAG

EtherType

S-TAG

EtherType

SA

DA

Provider Bridge IEEE 802.1ad

Payload

EtherType

C-TAG

EtherType

S-TAG

EtherType

SA

DA

I-TAG

EtherType

B-TAG

EtherType

B-SA

B-DA

Provider Backbone Bridge

IEEE 802.1ah

1990 1998 2005 2008


Provider Backbone Bridge (PBB) Terminology

PB = Provider Bridge BEB = Backbone Edge Bridge: inserts/removes the PBB header BCB = Backbone Core Bridge: similar behavior to 802.1ad bridge (aka QinQ)

BEB PB

BCB

BEB

PB

Customer

BCB


Quick Overview – PBB (con’t) I Component

• Maps S-VID to I-SIDs

• Adds PBB header without B-TAG

• Forwards frames to PB network based on customer MAC addresses

B-Component

• Maps I-SIDs to B-VIDs

• Adds B-TAG

• Forwards frames to core of PBB network based on backbone MAC addresses

BEB containing and I and B component is named IB-BEB

BEB can also support single component

• I-BEB

• B-BEB

Payload

EtherType

C-TAG

EtherType

S-TAG

EtherType

SA

DA

I-TAG

EtherType

B-TAG

EtherType

B-SA

B-DA

PBB IEEE 802.1ah

I Comp

B Comp


Quick Overview – PBB (con’t)

A

B

C

PBB Bridge Table

VID MAC Port

300 S20 4

I-SID Table 20

VID MAC Port

100 S1 3

100 A 3

100 B B-MAC S20

100 C B-MAC S20 S10

S20 S11

S1

S2

1 2

3 4

1

2

24

20 10

5

6 PBB Bridge Table

VID MAC Port

300 S10 10

I-SID Table 20

VID MAC Port

100 S2 20

100 A B-MAC S10

100 B 20

100 C 20

Bridge Table

VID MAC Port

300 S20 5

300 S10 6

Learn customer MACs only at edge nodes


Quick Overview – Shortest Path Bridging (SPB) Link state control plane for IEEE networks SPBV (Shortest Path Bridging – VID) / SPBM (Shortest Path Bridging – MAC with PBB)

Combines Ethernet Data Path (802.1Q or 802.1ah) with IS-IS (link state protocol)

Link State protocol used for (1) discovery, (2) advertise network topology, (3) compute shortest path trees from all bridges in the SPB Region

SPBV: Enables shortest path trees for VLAN Bridges

Defines a shortest path region, which is the boundary of the shortest path topology

Builds shortest path trees but also interworks with legacy bridges running rapid spanning tree protocol and multiple spanning tree protocol

SPBM: SPBM reuses the PBB data plane, which does not require that the Backbone Core Bridges (BCB) learn encapsulated client addresses

The forward and reverse paths used for unicast and multicast traffic in an IEEE 802.1aq network are symmetric

Equal Cost Multi Tree (path) supported (16 initially defined, more possible)


8 participating nodes MAC = 00:00:00:00:N:00 IS-IS runs on all the links Nodes will use their MAC addresses as IS-IS SysID to exchange link state packet (LSPs) After topology discovery the next step is distributed calculation of the unicast routes for both ECMP VIDs and population of the unicast forwarding tables (FIBs)

SPB - Example (1)

0 1

2 3

4

5

6

7

1 2 3

4

5

1

2

5

1 2 3

4

1 2

1

2 3

4

5

1 2

3 4 5

1 2

1

2


Node 7 will therefore have a FIB that among other things indicates: MAC 00:00:00:05:00 / VID 101 the next hop is interface/1. MAC 00:00:00:05:00 / VID 102 the next hop is interface/2 Node 5 will have exactly the inverse in its FIB. MAC 00:00:00:07:00 / VID 101 the next hop is interface/1. MAC 00:00:00:07:00 / VID 102 the next hop is interface/2 Equal Cost paths supported

SPB - Example (2)

0 1

2 3

4

6

7 5

1 2 3

4

5

1

2

5

1 2 3

4

1 2

1

2 3

4

5

1 2

3 4 5

1 2

1

2

Low path ID using VID 101

High path ID using VID 102


Leverages 802.1ag

SPB OAM Capabilities

Continuity Check (CC)

Fault detection (Multicast/unidirectional heartbeat)

Loopback – Connectivity check

Fault verification (unicast/bi-directional request/response)

Traceroute (link trace)

Fault isolation (trace nodes in path to a specified node)

Discovery (Y.1731/802.1ab)

Service (all nodes supporting common service instance)

Network (all devices common to a domain)

Performance Monitoring (MEF10, MEF12, Y-1731)

Capacity planning

SLA Reporting


SPB – Bottom line • Developed for service providers/carriers in the context of Internet L2 exchanges,

Metro Ethernet, Wireless Backhaul • SPB is actively supported by Alcatel Lucent, Huawei, Avaya (ex NT) and Ciena for DC

& DC to DC deployments • Leverages the industry standard Ethernet data planes – 802.1Q and 802.1ah • Supports tens of thousands of services with the 802.1ah I-SID (data path) • Leverages IS-IS link state protocol – already deployed by service providers/carriers

• Multiple shortest equal cost paths for both unicast and multicast traffic L2 VPNs

• Leverages the industry standard Ethernet OAM – 802.1ag


IETF: TRILL


TRILL: Introduction

A network where RBridges can Route packets to their target LAN. The paths they find, to our elation, Are least cost paths to destination! With packet hop counts we now see, The network need not be loop-free! draft-ietf-trill-rbridge-protocol-16 Ray Perlner, Algorhyme v2


TRILL - Terminology RBridge – Routing Bridges

• Benefits of both bridges and routers

• Terminates STP

• Invisible to IP routers

• Limited to customer networks

802.1

802.1

ES1 ES2

RBridge

End-station

Router

Ingress Egress Transit

IRB TRB ERB

Campus – TRILL network

• RBridges, bridges, hubs/repeaters (802.3)

• Bounded by end stations and routers

• Replaces old bridged LAN


TRILL in Action

Rbridges run IS-IS Link State protocol between each other

Optimal path found to every Rbridge

Small FDB for RBridge forwarding (100’s of RBridges)

Normal learning for end-stations (or ESADI protocol) Local MAC/VLAN/port

Remote MAC/VLAN/Rbridge Confidence level

Distribution trees for multicast (MCast, BCast, Unknown-uni) Pruned by VLAN

Pruned by IP Multicast membership


2-byte nicknames for Ingress and Egress RBridges Hop Count, Options length, Flags Routes packet within TRILL campus Transit switches do 16-bit lookup and decrement hop count If hop count is 0, packet discarded

TRILL Packet Format

Two headers added to original Ethernet packet Outer MAC header

TRILL header

Original packet excludes CRC

Total 20 bytes added

Outer MAC header

TRILL header

Original packet

18

6

60-1514

CRC 4

Outer DA

Outer SA

Outer VLAN

6

6

4

TRILL Etype 2

Needed for compatibility with 802.1 switches Outer DA is MAC address of next hop RBridge Outer SA is MAC address of sending RBridge Transit switches rewrite outer MAC header, like routers Outer VLAN is Etype (0x8100) and Designated VID Trill Etype indicates that 6-byte TRILL header follows

Outer Mac header

Hop Count, Flags

Egress RBr NN

Ingress RBr NN

2

2

2

TRILL Header


TRILL – Distribution Trees

• Used to forward multicast frames (Multicast/Broadcast/Unknown) • One tree is sufficient, but multiple trees allow load balancing • Tree computed based on link state information for a given root • All RBridges use LSPs to agree on: − Number of trees to compute

− Root of tree to compute

• RPF check protects against looping of multicast frames


TRILL – Distribution Trees (con’t)

Edge

Aggregation

Core

- TRILL RBridge

- IEEE 802.1 Switch

• IS-IS (Intermediate System to Intermediate System) link state routing protocol − IS-IS runs directly at Layer 2

− Optimal paths found between RBridges

25


IETF TRILL Forwarding Broadcast - TRILL RBridge

A B

Ethernet Frame IEEE 802.3

Ethernet Frame IEEE 802.3

TRILL

First communication host a send arp request to resolve host c mac

1 2

A->broadcast

S1 S2

S31 S30 S21

MAC Interface

A 1 S20 add MAC address of host A into MAC table

MAC DA is broadcast S20 will flood packet

S20 A->broadcast S20 -> broadcast

L1

L2

L4

L3

L5

L6 L7

L8

Switch Interface

S1 L1

S2 L2

S21 L1, L5

S30 L1, L5

S31 L1, L5

Switch Interface

S20 L1

S21 L2

S30 L3

S31 L4

S1 will flood frame based on local routing table


S31 decap header and flood frame

A->broadcast

L1 L2


IETF TRILL Forwarding Unknown unicast

- TRILL RBridge

27

A B

Host c will send arp reply back to host a

1 2

C->A

S1 S2

S31 S30 S21

MAC Interface

C 2

S31 MAC DA lookup will fail, frame will be flood

S20 C->A Broadcast -> S1

L1

L2

L4

L3

L5

L6 L7

L8

Switch Interface

S1 L10

S2 L9

S20 L9, L10

S21 L9. L10

S30 L9, L10

Switch Interface

S20 L1

S21 L2

S30 L3

S31 L4


L9 L10

S1 flood frame based on local table MAC Interface

A 1

C S31

Switch Interface

S1 L1

S2 L2

S21 L1, L2

S30 L1, L2

S31 L1, L2

S20 flood frame based on local table

S20 will see that A is already learned and will add C in local MAC table then decap header

C->A L1 L2


IETF TRILL Forwarding Unicast

- TRILL RBridge

28

A B

Host A starts sending traffic after arp resolution

1 2

A->C

S1 S2

S31 S30 S21

MAC Interface

A 1

C S31

Mac of C is in S20 table, encap frame

S20 A->C S20->S31

L1

L2

L4

L3

L5

L6 L7

L8

Switch Interface

S1 L1,

S2 L2

S21 L1, L2

S30 L1, L2

S31 L1, L2

S20 A->C S20 -> S31

L9

L10

S20 runs ECMP hash to select path MAC Interface

C 2

A S20

Switch Interface

S31 L9

S30 L7

S21 L6

S20 L5

S2 lookups S31 and pointed to L9

A->C


Determining which one is right for you


Decision criteria

Scalability Both technologies take care of the MAC explosion in the core

SPB with I-SID: 16M services. Edge encapsulation with TRILL can scale similarly

Failure recovery IS-IS computation may be faster with TRILL than with SPB – but only final implementations will

provide the real answers (highly debated issues)

Loop prevention

Loop mitigation

Both standards provide solutions (SPB: do not forward to root/agreement protocol. TRILL: TTL, RPC,

Adjacencies check)

Multicast Another highly debated issue. SPB: SPT calculated on every ingress node, more computational

intensive. TRILL: typically no more than 6 trees, simpler with fewer trees

Data center bridging TRILL: still work in progress (new draft). SPB: supported today

Compatibility 802.1D 802.1 bridges part of the TRILL domain. SPB a mode for normal VLAN bridges (V-mode)

OAM PBB/SPB leverage Ethernet OAM. Work in progress for TRILL (new draft)

https://datatracker.ietf.org/doc/draft-ietf-trill-rbridge-oam/


HP’s recommendations


IRF meets modern requirements today

HP 12500 Optimized network core Up to 4-chassis IRF available now

HP 5900/5920/58XX Optimized access layer 10/40 GbE Access

80% faster vMotion

500x faster recovery time

100% higher scalability

50% device reduction

20% lower price per port

300% higher scalability

Support for 1,000’s of virtual/physical servers

Resilient Virtual Switching Fabric IRF

Rack servers Blades servers


IRF / TRILL Comparison Scalability

IRF alone: Both edge and core switches learn the customer FDB

TRILL: core switches don’t learn edge MAC addresses

Failure recovery IRF: failover typically <100ms (link failover < 1ms)

TRILL: failover will depend on the implementation Loop prevention

Loop mitigation

IRF: part of the framework

TRILL: TTL & RPC & Adjacencies Check

Multi-pathing IRF: leverages ASIC hashing algorithms (L2/L3/L4)

TRILL: not specified by the standard, but expect 8 paths in first implementations

Data center bridging IRF: completely orthogonal property

TRILL: still work in progress (new draft)

Compatibility 802.1D IRF: does not require STP/RSTP/MSTP

TRILL: 802.1 bridges part of the TRILL domain

OAM IRF: specific OAM

TRILL: Work in progress for TRILL (new draft)

https://datatracker.ietf.org/doc/draft-ietf-trill-rbridge-oam/


IRF/TRILL Comparison (con’t)

• IRF and TRILL don’t play in the same dimension • IRF must be seen as a “clustering” technology allowing multiple devices to be

seen as one logical device, removing STP, VRRP from the network, with a single IP for the management

• TRILL answers positively to the following question: why can’t every single node have a tree rooted at itself, allowing (1) the optimal (shortest path) distribution of traffic (2) multi-pathing (3) failure recovery

• IRF and TRILL are in fact not mutually exclusive

34


HPN IRF/TRILL Data Center Fabric

Complementing TRILL with IRF

35

TRILL without IRF TRILL with IRF

• High performance : unblocking CLOS network

• Loop free, no STP

• 16 core switches, >100 10G boxes, >500 GE boxes

• Support or more than 20K servers

• Routing nodes >600

• 100% standardized TRILL, fully interoperable

• IRF reduces routing protocol (IS-IS) table size

• With 30 IRF domains (4 chassis per domain, 9

boxes at the edge)

• Only 30 routing nodes

• Allows larger domains, faster failure recovery

Combines best of both worlds !!!


Conclusion


Conclusion

Available today HPN’s IRF technology allows large DC deployments today

Active/Active links, L2 or L3, no STP/RSTP/MSTP (or VRRP)

HP is committed to TRILL Roadmap (POR)

Comware v7 / H2 2012

HP is committed to SPB PBB available today (12500/9500)

Roadmap (POI) Comware v7 / H1 2013

IRF + TRILL

Differentiated solution

Combining best of both worlds

Scalability, Faster Convergence, Ease of Use


Tools to help our clients • Read about the FlexNetwork Architecture

• Learn about Virtual Application Networks

• Discover Intelligent Management Center

• Read more on FlexFabric

• See more about FlexCampus BYOD for education and healthcare

• Learn how to simplify communication with FlexBranch

• View the HPN Portfolio Matrix Guide

• Learn about networking services from HP Technical Services

• Learn about networking career certifications from HP ExpertONE

http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA3-6037ENW.pdf


http://h17007.www1.hp.com/docs/interop/HP_Networking_Intelligent_Management_Center_Standard_Software_Platform_Datasheet_4AA3-0694ENW.pdf






http://www8.hp.com/us/en/services/services-detail.html?compURI=tcm:245-846146

http://www.hp.com/certification/expert_one-networking.html


Thank you