PBB-TE tests Victor Olifer (JANET/GEANT JRA1 Task 1) JRA1 Workshop, - - PowerPoint PPT Presentation

connect • communicate • collaborate

PBB-TE tests

Victor Olifer (JANET/GEANT JRA1 Task 1) JRA1 Workshop, Copenhagen, 20th November

connect • communicate • collaborate

Agenda

PBB-TE against EoMPLS History of trials JANET local trial UK-wide testbed & PBB-TE and EoMPLS interworking tests Testing PBB-TE resilience (protection switching) General conclusions of EoMPLS & PBB-TE trial

connect • communicate • collaborate

Intro: technologies & features Two reps of Carrier Ethernet:

• Two-tier hierarchy
• Traffic Engineering
• Protection switching
• Ethernet&MPLS OAM
• Two-tier hierarchy
• Traffic Engineering
• Protection switching
• Ethernet OAM
• Established
• Rich control plane
• Complex
• Multi-domain support:
• Local labels
• BGP
• Emerging
• Zero control plane
• Simple (relatively)
• Single-domain:
• Global labels
• GMPLS? Not avail.

connect • communicate • collaborate

Carrier Ethernet family objectives

De-coupling of provider and user networks PB – VLAN ID separation PBB/PBB-TE – MAC and VLAN ID separation Resilience PB & PBB – STP (TRILL, SPB) – re-routing PBB-TE – fast protection switching Traffic Engineering PB, PBB – no (and yes for non-resilient services if routing is switched off – VLAN- based path ) OAM Relevant for PB, PBB, PBB-TE

connect • communicate • collaborate

History of trials

2008 2009 2010 2011

JANET Local trial JANET UK-wide Carrier Ethernet trial JRA 1 Task 1 PBB-TE trial

same testbed

connect • communicate • collaborate

PBB-TE local trial

Switch A CIENA 311v Switch B CIENA 311v Switch C CIENA 311v

Simple goal: To check whether this new Carrier Ethernet offspring does what his parents promise Results: In general: Yes, it does, and in a very familiar to classic Ethernet way

• TE – yes, by establishing of PBB-TE tunnels with explicit path
• Scalability – yes, by using customer (I-SID) connections over tunnels

up to 16 M connections per tunnel

MAC A MAC C B‐VID N B‐VID N B‐VID N B‐VID N

connect • communicate • collaborate

MACinMAC encapsulation

Customer network

Customer network PB network PBB/PBB TE network PB network S‐VID added B‐header added B‐header removed S‐VID removed

connect • communicate • collaborate

PBB-TE local trial (cont.)

CIENA 311v CIENA 311v CIENA 311v

Results:

• Resilience – yes, by fast protection switching
• f tunnels triggered by CCM heartbeat messages

Primary tunnel Backup tunnel

connect • communicate • collaborate

Warrington Reading London Telecity

Core

Manchester Uni Oxford Uni Lancaster Uni Essex Uni

JANET/JRA1 Task 1 Carrier Ethernet multi-domain testbed

JANET Lightpath (EoMPLS)

• PBB-TE domain
• EoMPLS-domain

JANET(UK)/Lumen House CIENA 311v CIENA 311v CIENA 311v

connect • communicate • collaborate

Warrington Reading London Telecity

Core

Manchester Uni Oxford Uni Lancaster Uni Essex Uni

PBB-TE & EoMPLS interworking tests:

• 1. EoMPLS – PBB-TE – EoMPLS

JANET Lightpath (EoMPLS)

MEF E-NNI: S-VID (outer VID) – service delimiter

M-Eth SA M-Eth DA MPLS LSP MPLS PW C-Eth DA C-Eth SA

Payload (IP)

B-SA B-DA B-VID I-SID C-DA C-SA

Payload (IP)

S-VID

• 1. Use tagged Ethernet frames and copy/map PW ID into S-VID
• 2. Encapsulate EoMPLS frames into PBB-TE frames at ingress
• 3. Copy/map S-VID into I-SID
• 4. De-capsulate EoMPLS frames at ingress and send to

destination

Payload M-Eth SA M-Eth DA MPLS LSP MPLS PW C-Eth DA C-Eth SA

Payload (IP)

S-VID B-SA B-DA B-VID I-SID C-DA C-SA

Payload (IP)

Payload M-Eth SA M-Eth DA MPLS LSP MPLS PW C-Eth DA C-Eth SA

Payload (IP)

S-VID M-Eth SA M-Eth DA MPLS LSP MPLS PW C-Eth DA C-Eth SA

Payload (IP)

S-VID

connect • communicate • collaborate

PBB-TE & EoMPLS interworking tests:

• 1. EoMPLS – PBB-TE – EoMPLS (cont.)
• Overlay mode for the core, conforms to MEF E-NNI
• Contiguous MPLS tunnels and PWs
• Usage of IP control plane protocols in the EoMPLS testbeds

(partly to make it close to real JANET):

• OSPF, BGP, LDP, RSVP (only for TE)

Main characteristics of the solution: Problems encountered:

• STP BPDUs received within MPLS PWs from neighboring MPLS domain

confused local STP and resulted in blocking ports :

• It was fixes by switching STP off
• LDP refused to distribute labels between MPLS domains which

belonged to different AS:

• It was fixes by using ‘BGP send-label’

connect • communicate • collaborate

PBB-TE & EoMPLS interworking tests:

• 2. PB – PBB-TE – EoMPLS- PB

Warrington London Telecity

Core

Essex Uni JANET(UK)/Lumen House JANET Lightpath (EoMPLS)

Two modes for the core and peripheral testbeds were tested:

1. Overlay, with PB in the peripheral testbeds and encapsulation into PBB-TE in the core For LH – Essex Uni connection:

• 1. PB frame is encapsulated into PBB-TE one at

the core ingress E-NNI in Reading

Reading B-SA

B-DA B-VID I-SID

Payload (IP)

C- SA C- DA C-VID

Payload (IP)

S-VID

E-NNI

• 2. S-VID is copied/mapped into I-SID
• 3. PBB-TE frame travels to the egress at Telecity

switch using I-SID as a service delimiter

B-SA B-DA B-VID I-SID

Payload (IP)

C- SA C- DA C-VID

Payload (IP)

S-VID

• 4. PB frame is de-capsulated at the core egress

C- SA C- DA C-VID

Payload (IP)

S-VID

• 5. PB frame is delivered to Essex Uni testbed

through JANET Lightpath EoMPLS connection

One more overlay transfer: PB over EoMPLS on basis of S-VID

connect • communicate • collaborate

PBB-TE & EoMPLS interworking tests:

• 2. PB – PBB-TE – EoMPLS- PB (cont.)

Warrington London Telecity

Core

Essex Uni JANET(UK)/Lumen House JANET Lightpath (EoMPLS)

Second mode tested:

Peer-to-peer mode with a contiguous PBB-TE connection For LH – Essex Uni connection:

• 1. Customer frame is encapsulated into PBB-TE
• ne at the LH testbed ingress UNI

Reading

B-VID B-SA B-DA I-SID

Payload (IP)

C- SA C- DA C-VID

Payload (IP)

I-NNI

• 2. C-VID is mapped into I-SID
• 3. PBB-TE frame travels along the contiguous

PBB-TE tunnel (LH – the core – Essex Uni) using I-SID as a service delimiter

• 4. PB frame is de-capsulated at Essex testbed

egress UNI

C- SA C- DA C-VID

Payload (IP)

B-VID B-SA B-DA I-SID

Payload (IP)

C- SA C- DA C-VID

Payload (IP)

PBB-TE frame travelled over Lightpath EoMPLS on basis of B-VID (outer VID)

connect • communicate • collaborate

PBB-TE & EoMPLS interworking tests:

• 2. PB – PBB-TE – EoMPLS- PB (cont.2)

Overlay PB vs. contiguous multy-domain PBB-TE

Overlay model Contiguous model

Number of PBB-TE tunnels Minimal:

• Only to connect

domain edge switches (e.g. 3 unprotected core tunnels in our case) Might be quite big:

• A tunnel per

customer pair (e.g. 10 unprotected core tunnels in our case) Co-ordination of end point of tunnel MAC addresses between domains Not needed Needed (private loopback MACs might be used) Tunnel protection Only within a domain End-to-end IP control plane Not needed, doesn’t exist yet in practice (might be GMPLS) Not needed, doesn’t exist yet in practice (might be GMPLS)

connect • communicate • collaborate

Warrington Reading London Telecity

Core

Lancaster Uni Essex Uni

Overlay and contiguous PBB-TE protection switching

JANET LH

• 1. Overlay model

Primary tunnel Backup tunnel

X

No mechanism to redirect traffic in case of inter-domain link failure – so, only intra-domain protection Control Plane inter-domain protocol is needed – e.g. BGP

connect • communicate • collaborate

Warrington Reading London Telecity

Core

Lancaster Uni Essex Uni

Overlay and contiguous PBB-TE protection switching (cont.)

JANET LH

• 2. Contiguous model

Primary tunnel Backup tunnel

Standard CCM mechanism triggers end-to-end protection No other Control Plane inter-domain protocol is needed

X

connect • communicate • collaborate

General Carrier Ethernet trial conclusions

• Both EoMPLS and PBB-TE proved to be working transport

technologies with required core set of carrier-grade features

• EoMPLS and PBB-TE can smoothly inter-operate according MEF E-NNI spec
• EoMPLS is a good choice for carrier core networks because of its tight

integration with powerful IP control plane, router vendor support and wide implementation base

• PBB-TE might be used for access and campus networks: simple but robust
• PBB-TE is not dead despite some rumours (Ciena, Extreme, ...)

connect • communicate • collaborate

Trials’ participants

 JANET Carrier Ethernet Trial

Victor Olifer (JANET UK)

(victor.olifer@ja.net)

Dave Tinkler (JANET UK) Martin Dunmore (JANET UK) Michael Robson (Manchester Uni) Anthony Ryan (Manchester Uni) Faris Ali (Lancaster Uni) Oliver Gorwitz (Oxford Uni) Guy Morrell (Oxford Uni) Bijan Rahimzadeh Rofoee (Essex Uni)

 JRA1 Task 1:

Jan Radil (CESNET) Marcin Grastka (PSNC) Ramanujam Jayakumar (Essex Uni) Jac Kloots (SURFnet) Alberto Colmenero (NORDUnet)

connect • communicate • collaborate

PBB-TE positioning

Wavelength and sub wavelength switching: DWDM/OTN/GFP Sub wavelength switching: SDH Frame switching: PBB-TE Packet switching: IP/MPLS; Services

connect • communicate • collaborate

PBB-TE positioning (cont.)

Wavelength and sub wavelength switching: DWDM/OTN/GFP Packet switching: IP/MPLS; Services Frame switching: Carrier Ethernet

• 1. L2 services to customers
• 2. Links to upper layers
• 3. Links to upper layers directly from layer 0/1

connect • communicate • collaborate

PBB-TE - Optical integration

Wavelength, sub wavelength and frame switching: DWDM/OTN/GFP/CE

• Pros
• Cons
• Network is simpler: one layer, less boxes, one type of boxes
• More efficient provisioning :

consistent approach of one NMS or one control plane -> cut through, selection between layers etc Complexity of combined boxes –>

• difficulty in configuring grows as the number of

components are squared -> error prone, unstable behaviour

connect • communicate • collaborate

Questions

Questions?

connect • communicate • collaborate

Extra slides

connect • communicate • collaborate

Testing PBB-TE resilience

warr‐cec1 Essex Lancaster

• The configuration was tested for a ping session with 100 ms interval

PBB‐TE tunnel group: warr‐lond

7/2 7/1 7/2 7/1 7/23 7/24 7/23 7/24 7/24

VS: lancaster‐essex‐vs

read‐cec1 lond‐cec3

VS: essex‐lh‐vs Tunnel 1 : weight 8 Tunnel 2 : weight 7 Tunnel 3 : weight 6

X X

• CCM inetrval was set also for 100 ms
• Switching off port 7/1 of warr‐cec1 caused loss of 0 or 1 ping

7/1 7/2

connect • communicate • collaborate

Traffic policing tests