Metro Ethernet: Understanding Key Underlying Technologies

Download Report

Transcript Metro Ethernet: Understanding Key Underlying Technologies

Metanoia, Inc.
Critical Systems Thinking™
Metro Ethernet:
Understanding Key Underlying
Technologies
Metanoia, Inc.
[email protected]
+1-888-641-0082
http://www.metanoia-inc.com
© Copyright 2007
All Rights Reserved
Metanoia, Inc.
Critical Systems Thinking™
Who is Metanoia, Inc.?

Specialty technology consultancy founded in mid-2001, with HQ in Mountain View, California

Undertakes deep-dive technical & strategy consulting in telecom network, systems, software and
chip architecture and design for clients across the world

Services have spanned 4 continents, with clients in: North America, Europe, Asia, and Australia.

Principals provided services in technology strategies, architecture and design trade-offs, product
development, hardware/software architecture, and knowledge enhancement to organizations that
include large equipment manufacturers, international, national and regional ISPs, premier metro/access
systems startups, network planning tool vendors, established software and technology houses and
leading component and semiconductor vendors

Principals are technologists at the forefront of new developments, as leaders, creators,
implementers, researchers, academics, strategists, and advisors in the US and abroad

Expertise spans Layer 1 through Layer 4, and wireline (optical, Ethernet, IP/ATM, SONET/SDH)
through wireless (Wi-Fi, cross-layer design, Wi-Max, cellular data, 2.5-3G, LTE)

125+ man years of technology design and development, and technology management experience,
having worked/consulted at leading global corporations, such as Apple, AOL Time Warner, BBN, Cisco,
3Com, Fujitsu, LSI Logic, Motorola, Tellabs, Siemens, Nokia, Tibco, and Qualcomm, and having worked
at/consulted to corporates in the US and abroad for almost the last decade

70+ patents collectively issued/pending

Advanced graduate degrees from some of the most distinguished universities in the world – the
University of California, Stanford University, Iowa State University, and the Indian Institute of
Technology
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
2
Metanoia, Inc.
Critical Systems Thinking™
Workshop Outline
 Legacy networks & Ethernet over legacy networks
 Value propositions and business drivers
 Ethernet over SDH/SONET
 Metro Ethernet Forum (MEF)
 MEF architecture

E-Line and E-LAN services
 Native Ethernet as Carrier-class transport
 Provider Bridges
 Provider Backbone Bridges (PBB), Provider Backbone Transport (PBT)
 MPLS – an enabler for Ethernet services
 Layer 2 VPNs: VPWS, VPLS, H-VPLS
 Advanced concepts: traffic engineering, QoS, OAM, resilience
 Conclusions
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
3
Metanoia, Inc.
Critical Systems Thinking™
Ethernet over
Legacy Networks
Metanoia, Inc.
Critical Systems Thinking™
Issues with Legacy Networks
 Low bandwidth
 No flexibility to scale
 High cost of installation
 Slow provisioning
 Bandwidth growth inflexible/non-linear
 Limited by multiplexing hierarchy
 TDM-based access: inefficient for converged data
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
5
Metanoia, Inc.
Critical Systems Thinking™
Next-Generation SDH
Customer
Network
Central
Office
Switch
NG-SDH
NG ADM
Core
Network
NG-SDH
Ethernet
NG ADM
Cross
Connect
Customer
Network
STM/4/16
Ring
NG NG-SDH
ADM
Ethernet
Customer
Network
Customer
Network
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
6
Metanoia, Inc.
Critical Systems Thinking™
Ethernet-over-SDH
 Framing protocol
 Encapsulates Ethernet frames in SDH payloads
 Mapping of SDH payload to SDH channels
 Virtual concat.: for allocation of non-contiguous VCs
 Flow control mechanism
 Avoids packet drops due to speed mismatch between SDH and
Ethernet
 Mechanism to increase/decrease allocated SDH bandwidth
 Add or remove VCs
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
7
Metanoia, Inc.
Critical Systems Thinking™
Ethernet-over-SDH (contd)
 Very popular in carriers with installed base of SDH rings
 E.g. BSNL in India
 Good deployment choice when traffic primarily circuit
switched
 Inefficient if major traffic is bursty packet-switched data
 Solution: Carrier-class Ethernet!
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
8
Metanoia, Inc.
Critical Systems Thinking™
Metro Ethernet Value Propositions
 Lower per-user provisioning costs
 Technically simple relative to TDM ckts.
 Due to large installed base
 Efficient and flexible transport
 Wide range of speeds: 128 Kbps--10 Gbps
 QoS capabilities
 Ease of inter-working
 Plug-and-play feature
 Ubiquitous adoption
 The technology of choice in enterprise networks
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
9
Metanoia, Inc.
Critical Systems Thinking™
Ethernet Business Drivers
 Business connectivity
 Storage networks
 Data centers
 Video conferencing
 Residential services
 Triple-play services (IPTV)
 On-line gaming
 High-speed Internet access
 Wireless backhaul
 Reduced cost, complexity for mobile operators
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
10
Metanoia, Inc.
Critical Systems Thinking™
Metro Ethernet Services
Metanoia, Inc.
Critical Systems Thinking™
Metro Ethernet Forum (MEF)
 Industry forum at forefront of Carrier Ethernet
standardization
 Carrier Ethernet architecture
 Ethernet services
 Founded in 2001. Currently approx. 120 members
 Technical Sub-committees
 Architecture
 Services
 Protocols and Transport
 Management
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
12
Metanoia, Inc.
Critical Systems Thinking™
MEN Architectural Components
T
T
S
S
End
User
Customer
Network
Customer
Network
MEN
End user Interface
UNI Reference Point
End
User
End user Interface
UNI Reference Point
Ethernet Virtual Connection
End-to-End Ethernet Flow
 Ethernet Flow
 Unidirectional stream of Ethernet frames
 UNI
 Interface used to interconnect MEN subscriber to provider
 EVC
 Defines association between UNI for delivering Ethernet flow across MEN
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
13
Metanoia, Inc.
Critical Systems Thinking™
MEN Layer Model
Application Service
Layer
(IP, MPLS, PDH, E1/E3, SDH)
Ethernet Service
Layer
Transport Service
Layer
(802.1, SONET/SDH, MPLS)
MEN Layer Model
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
14
Metanoia, Inc.
Critical Systems Thinking™
MEF Services Definition Framework
 Service Type
 Construct used to create broad range of services
 Service Attributes
 Defines characteristics of a service type
 Attribute Parameters
 Set of parameters with various options
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
15
Metanoia, Inc.
Critical Systems Thinking™
Service Types
 E-Line
 Point-to-point Ethernet Virtual
EVC1
Circuit (EVC)
EVC2
 E-LAN
 Multipoint-to-multipoint
Ethernet Virtual Circuit
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
16
Metanoia, Inc.
Critical Systems Thinking™
Service Attributes
 Physical Interface
 Medium, speed, mode, MAC layer
 Traffic Parameters
 CIR, CBS, PIR, MBS
 QoS Parameters
 Availability, delay, jitter, loss
 Service Multiplexing
 Multiple instances of EVCs on a given physical I/F
 Bundling
 Multiple VLAN IDs (VID) mapped to single EVC at UNI
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
17
Metanoia, Inc.
Critical Systems Thinking™
Ethernet Services
 Ethernet Private Line (EPL)
 Uses E-Line
 Does not allow service multiplexing
 High degree of transparency
 Low delay, delay variation, and packet loss ratio
 Ethernet Virtual Private Line (EVPL)
 Uses E-Line
 Allows for service multiplexing
 Need not provide full transparency
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
18
Metanoia, Inc.
Critical Systems Thinking™
Service Types and Ethernet Services
Service Types
E-Line
(p2p connectivity)
Ethernet Private
Line (E-line)
Ethernet Virtual
Private Line (E-VPL)
E-LAN
(mp2mp connectivity)
Ethernet Private
LAN (E-LAN)
Ethernet Virtual Private
LAN (E-VPLAN)
Ethernet Services
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
19
Metanoia, Inc.
Critical Systems Thinking™
Native Ethernet as
Carrier-class Transport
Metanoia, Inc.
Critical Systems Thinking™
Requirements for Carrier-class Ethernet
 Scalability
 Network should support millions of subscribers
 Protection and restoration
 50ms resilience
 Quality-of-Service (QoS)
 Ability to offer differentiated levels of service
 Service Monitoring and Fault Management
 Support for TDM traffic
 Seamless integration with legacy networks
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
21
Metanoia, Inc.
Critical Systems Thinking™
Ethernet Ring
Ethernet
Switch
Ethernet
Switch
Core
Network
Ethernet
Switch
Ethernet
1/10 Gigabit
Ethernet Ring
Customer
Network
Ethernet
Switch
Ethernet
Copyright 2007
All Rights Reserved
Customer
Network
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
22
Metanoia, Inc.
Critical Systems Thinking™
Native Ethernet in Metro Access
 How does one create the notion of a virtual circuit?
 VLAN tagging with point-to-point VLAN
 VLAN stacking
 Outer tag  service instance; Inner tag  individual customer
 802.1Q in 802.1Q (Q-in-Q) - IEEE 802.1ad
6bytes
C-DA
6bytes
C-SA
4bytes
4bytes
S-TAG
C-TAG
4bytes
Client data
FCS
C-DA: Customer Destination MAC
C-SA: Customer Source MAC
C-TAG: IEEE 802.1q VLAN Tag
C-FCS: Customer FCS
S-TAG: IEEE 802.1ad S-VLAN Tag
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
23
Metanoia, Inc.
Provider Bridge (IEEE 802.1ad)
Architecture
Critical Systems Thinking™
CE-B
CES
CE-A
UNI-B
Customer
Network
Customer
Network
CES
UNI-A
CES
Spanning tree
UNI-C
CE: Customer Equipment
UNI: User-to-Network Interface
CES: Core Ethernet Switch/Bridge
CE-C
Customer
Network
P-VLAN: Provider VLAN
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
24
Metanoia, Inc.
Critical Systems Thinking™
Limitations of Provider Bridge Scalability
 Limited to 4096 service instances
 Core switches must all MAC addresses
 Broadcast storms ensue due to learning
 MAC address tables explode!
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
25
Metanoia, Inc.
Critical Systems Thinking™
Provider Backbone Bridging (802.1ah)
 Encapsulate customer MAC with provider MAC at edge
 Edge switch adds 24-bit service tag (I-SID), not VLAN tag
 Core switches need only learn edge switch MAC adds.
6bytes
6bytes
4bytes
B-DA
B-SA
B-TAG
5bytes
I-TAG
6bytes
6bytes
4bytes
C-DA
C-SA
C-TAG
4bytes
Client data
B-FCS
S-TAG: IEEE 802.1ad S-VLAN Tag
B-DA: IEEE 802.1ah Backbone Destination
B-SA: IEEE 802.1ah Backbone Source MAC
I-TAG: IEEE 802.1ah Service Tag
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
26
Provider Backbone Bridging (PBB)
Architecture
CPE A
CPE B
Provider backbone
network (802.1ad)
CPE C
CPE A
Metanoia, Inc.
Critical Systems Thinking™
CPE B
CPE D
Provider backbone
network (802.1ad)
802.1ad
Provider backbone
network (802.1ah)
Provider backbone
network (802.1ad)
Provider backbone
network (802.1ad)
802.1q
CPE C
CPE B
CPE B
CPE A
CPE D
CPE C
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
27
Metanoia, Inc.
Critical Systems Thinking™
Benefits of PBB
 Scalability
 Addresses limitations of 4096 service instances
 Robustness
 Isolates provider network from broadcast storms
 Security
 Provider need switch frames only on provider addresses
 Simplicity
 Provider & customers can plan networks independently
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
28
Metanoia, Inc.
Critical Systems Thinking™
Traffic Engineering in PBB
 Via Multiple Spanning Tree Protocol (MSTP)
 Maps a VLAN to ST or multiple VLANs to ST
 Enables use of links that would otherwise be idle in ST
 Eliminates wasted bandwidth … but …
 Too slow for protection switching
 Not suitable for complex mesh topologies
 Difficult to predict QoS
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
29
Challenges with an All-Ethernet
Metro Service
Metanoia, Inc.
Critical Systems Thinking™
 Restriction on # of customers – 4096 VLANs!
 Service monitoring
 Scaling of Layer 2 backbone
 Service provisioning
 Carrying a VLAN is not a simple task!
 Inter-working with legacy deployments
 Need hybrid architectures …
Multiple L2 domains connected via IP/MPLS backbone
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
30
Metanoia, Inc.
Critical Systems Thinking™
What Solutions do we Have?
 Ethernet-based Architecture
 Provider Bridge (802.1ad) in edge
 Provider Backbone Transport (PBT) in Core
 Hybrid Architecture
 802.1ad in the edge
 Multiprotocol Label Switching (MPLS) in core
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
31
Metanoia, Inc.
Critical Systems Thinking™
Provider Backbone Transport (PBT)
 Connection-oriented, traffic-engineered Ethernet tunnels
 Replaces spanning tree control plane with either a:
 Management plane
 External control plane
 No learning !
 Forwarding info. provided by management plane
 Forwarding done on MAC + VID (60-bit) address
 VID is not network global; however, MAC + VID is
 B-MAC identifies destination
 B-VID identifies per-destination alternate paths
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
32
Metanoia, Inc.
Critical Systems Thinking™
PBT Architecture
Central TE Module
PE2
PE1
Customer
Network
Customer
Network
SA : PE1
DA : PE2
VLAN 22
Copyright 2007
All Rights Reserved
SA : PE1
DA : PE2
VLAN 33
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
33
Metanoia, Inc.
Critical Systems Thinking™
Benefits of PBT
 No learning
 Eliminates undesirable broadcast storms
 Resolves MAC flooding problem
 Addresses scaling by forwarding on MAC + VID-highly scalable
 Protection
 Sets-up backup paths
 50ms restoration possible
 QoS support available
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
34
Metanoia, Inc.
Critical Systems Thinking™
MPLS – An Enabler for
Ethernet Services:
Fundamentals & Operations
Metanoia, Inc.
Critical Systems Thinking™
Basic Concept of MPLS
Next hop
router
129.89.10.x 198.168.7.6
N/w
Int.
1
DA
179.69.x.x
1
DA
198.168.7.6
129.89.10.x
Next hop
router
129.89.10.1
N/w
Int.
1
179.69.x.x
179.69.42.3
2
Routing Table
128.89.10.x
In
label
Out
label
X
3
4
X
Address Prefix N/w
Int.
128.89.10.x
179.69.x.x
1
1
In
label
Out
label
3
4
5
128.89.10.x
1
7
179.69.x.x
2
Address Prefix N/w
Int.
1
R1
1
128.89.10.12
Label Table
R3
Advertises binding
<5, 128.89.10.x>
R2
198.168.7.6
Advertises bindings
<3, 128.89.10.x>
<4, 179.69.x.x>
2
Advertises binding
<7, 179.69.x.x>
 Routing fills routing table
 Signaling fills label forwarding table
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
179.69.x.x
R4
179.69.42.3
36
Metanoia, Inc.
Critical Systems Thinking™
Basic Concept of MPLS
In
label
Out
label
X
3
X
4
In
label
Out
label
1
3
5
128.89.10.x
1
1
4
7
179.69.x.x
2
Address Prefix N/w
Int.
128.89.10.x
179.69.x.x
Pop
label
5
Address Prefix N/w
Int.
5
Forward
packet
128.89.10.x
128.89.10.12
R3
Swap
Label
1
3
R1
1
R2
198.168.7.6
3
Packet arrives
DA=128.89.10.25
5
2
Push
Label
179.69.x.x
R3
R4
179.69.42.3
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
37
So what about MPLS Control and
Forwarding?
Metanoia, Inc.
Critical Systems Thinking™
 Superset of conventional router control
 Distribute info. via n/w layer routing protocols (OSPF, BGP, etc.)
Control
Component
 Algos. to convert routing info. into forwarding table:
 Create binding from FEC  label
 Assign & distribute labels to peer LSRs via signaling
 Label switching forwarding table (or label information base LIB)
Incoming Label
Map
Incoming
Label
Forwarding
Component
First Subentry
Outgoing label
Outgoing inf.
Next hop address
Second Subentry
(for multicast or load balancing)
Outgoing label
Outgoing inf.
Next hop address
Next hop label forwarding entry (NHFLE)
 Forwarding algo = label swapping, independent of control
component (implementable in optimized H/W or S/W)
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
38
What does a Label Represent? The
Issue of Label Granularity
Metanoia, Inc.
Critical Systems Thinking™
 Packets form Forwarding Equivalence Class (FEC)
 Treated identically by participating routers
 Assigned the same label
 Membership in FEC must be determinable from IP header + other info. that
ingress router has about the packet
 Entities that may be grouped into an FEC are flexible. E.g. FEC could be:
 Connection between two IP ports on two hosts or between IP hosts
 Traffic headed for a particular network with same TOS bits
 All destination networks with a certain prefix
 Manually configured connection
 Traffic belonging to a customer or department VLAN
 Traffic of a given application – voice, video, plain data, management traffic
… and many others
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
39
Metanoia, Inc.
Critical Systems Thinking™
Let’s Recap: Elements of MPLS
 Label Forwarding
 Use data link addressing. E.g. ATM VPI/VCI, FR DLCI
 “Shim” header between data link and IP header
Data
Plane
Variable
L2 header
4 bytes
20 bytes
MPLS “shim”
header
L3 IP header
Higher Layers
1 bit
Label
20 bits
EXP/
S
CoS
TTL
3 bits
8 bits
 Label Creation and Binding
Control
Plane
 Label Assignment and Distribution
 Ride piggyback on routing protocols, where possible (BGP)
 Separate label distribution protocol – RSVP, LDP
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
40
Primary Label Assignment and
Distribution Modes
1
Edge LSR
Metanoia, Inc.
Critical Systems Thinking™
Requests
2
6
5
3
4
Downstream-on-demand
with Ordered Control
Assignments
1
Edge LSR
Edge LSR
Requests
2
Assignments
2’
3’
Downstream-on-demand
with Independent Control
3
4
Edge LSR
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
41
Metanoia, Inc.
Critical Systems Thinking™
Advantages of MPLS
Original justification
 Availability of fast, amortized, ATM hardware; emergence of H/W
forwarding engines has practically eliminated this
Current justifications
 Separates forwarding from control, allowing
 Routing functionality to evolve independently of forwarding algorithm
 MPLS to control non-packet technologies: SONET/SDH ckts., lightpaths
 Provides explicit, manageable IP routes
 Enables policy routing and traffic engineering
 Offers TE for Ethernet tunnels in metro-Ethernet environments
 Facilitates scalable hierarchical routing
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
42
Metanoia, Inc.
Critical Systems Thinking™
The Utility of Hierarchical Label Switching
Edge LSRs
Swap
Swap
and Push
Core LSRs
Pop
Concept is similar to VLAN stacking in PBT we saw earlier
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
43
Metanoia, Inc.
Critical Systems Thinking™
Hierarchical Label Stacking/Switching
 Inside a transit AS, each core router must keep track of all
networks that might be reached through it
 With hierarchical labels, only edge routers need know what
networks might eventually be reached through them
 All transit traffic can be made to tunnel through core routers
using LSPs with stacked labels
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
44
Explicit Manageable Routes -- Policy
routing, Traffic engineering
Metanoia, Inc.
Critical Systems Thinking™
 Carriers want certain traffic to go over certain routes. Such
network engineering:
 Keeps network loads balanced
 Enhances network stability and reliability
 Enables better QoS and performance assurances
 Allows carriers to meet customer SLAs
 Constraint-based routing together with MPLS allows carriers to
 Bind Ethernet tunnels to an LSP,
 Place (or route) LSP over the desired sequence of LSRs in the n/w
 TE tunnels are helpful for VPLS-based carrier Ethernet n/ws
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
45
Metanoia, Inc.
Critical Systems Thinking™
IP/MPLS-based Layer 2 VPNs
Metanoia, Inc.
Critical Systems Thinking™
L2 VPN Components
VC LSP
A
A
PE1
Emulated
LAN A
PE2
Routed
backbone
B
B
AC
Emulated
LAN B
PE3
What does the P1-PE2
connection really look like?
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
47
Metanoia, Inc.
Critical Systems Thinking™
L2 VPN Component Details
6
PW Signaling
PE1
From CE
devices
PE2
5
PSN Tunnel
3
1
ACs
PWs
Routed backbone
with P routers
2
Bridge
Module
4
Forwarder
From CE
devices
Emulated LAN
Instance
Emulated LAN
Interface
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
48
Metanoia, Inc.
Critical Systems Thinking™
VPLS Network Overview
PW
(full mesh)
LAN Service
A
VSI
VSI
VSI
CE
L3/MPLS
Backbone
VSI
B
B
CE
AC
A
Copyright 2007
All Rights Reserved
VSI
Tunnel
(full mesh)
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
LAN Service
49
Metanoia, Inc.
Critical Systems Thinking™
VPLS Protocols Involved
Control Ethernet
Plane
STP
MP-iBGP (PW) + RSVP-TE /LDP (tunnel)
Targeted LDP (PW) + LDP (tunnel)
Ethernet
STP
A
PE
CE
BGP/Targeted LDP
PE
LSP or PSN Tunnel
B
B
CE
Data
Plane
Ethernet
Copyright 2007
All Rights Reserved
Ethernet or
Ethernet in IP/
ATM/FR/SDH/
SONET
Ethernet/MPLS
Ethernet/IPSec
Ethernet/GRE
Ethernet
Ethernet or
Ethernet in IP/
ATM/FR/SDH/
SONET
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
50
Metanoia, Inc.
Critical Systems Thinking™
Operational Characteristics of VPLS
Operational Requirement
Realized Via
MAC address learning and
switching, work with 802.1p/q
tags and VLANs
- VSI Forwarder
- Bridge Module
Flooding pkts. with unknowns
broadcast, or multicast address
Frame replication on PWs
Provider edge signaling – inform
- Targeted LDP
PE's to autoconfigure, and of
- BGP
membership, tunnelling
VPLS membership discovery
- BGP
- Configuration
Inter-provider connectivity
Globally unique VPLS ID
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
51
Metanoia, Inc.
Data Plane: Flooding, Address
Learning and Forwarding
Critical Systems Thinking™
Src. MAC = 09:10:01:45:00:AB
Dest. MAC = 08:00:69:02:01:FC
1
3
VSI
CE
2
PE2
PWs
PE1
B
VSI
?
VSI
A
2
PE3
A
VSI
PE4
B
VSI
CE
3
 All address unknown frames (unicast, multicast, broadcast)
flooded over corresponding PWs to all relevant PEs only
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
52
Metanoia, Inc.
Critical Systems Thinking™
Address Learning
 Layer 2 reachability directly learned in data plane
 Use standard learning bridge functions for local MACs
 PW-based association for remote MACs
 Allow PE to determine from which physical port or LSP a given MAC
address came
 VSI FIB keeps mapping between Ethernet MAC  PW to use
Qualified Learning
- Each customer VLAN is its own
VPLS instance
- Has its own PW mesh and brdcast
domain
Copyright 2007
All Rights Reserved
Unqualified Learning
- All customer VLANs are part of
the same VPLS
- One PW mesh and single brdcast
domain
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
53
Metanoia, Inc.
Critical Systems Thinking™
Address Learning Example
2
Src. MAC = 08:AA:FC:01:10:DE (S1)
Dest. MAC = FF:FF:FF:FF:FF:FF (D1)
(broadcast)
4
VSI
1
Inbound
VC LSP Label = 1002
CE
i/f2
i/f1
VSI
i/f1
PE1
PE2
3
Local Learning
Outbound
VC LSP Label = 2001
Dest. VC
Tunnel Out I/F
MAC Label
i/f1
1002
S1
PE3
Copyright 2007
All Rights Reserved
A
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
Remote
Learning
54
Metanoia, Inc.
Critical Systems Thinking™
Forwarding and Encapsulation
Forwarding requires ability to
 Dynamically learn MAC addresses on
 Physical ports
 Pseudowire VCs (VC LSPs)
 Forward/replicate pkts. across physical ports and VC LSPs
Encapsulation
 PW header applied to Ethernet packet w/o preamble + FCS
 VLAN tag denoting customer’s VPLS instance can be stripped at
ingress, reapplied at egress
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
55
Metanoia, Inc.
Tunnel and PW Topology and
Loop Freedom
Dest. MAC = 08:00:69:02:01:FC
VSI
?
Critical Systems Thinking™
PW
(full mesh)
A
VSI
PE2
PE1
VSI
VSI
CE
B
AC
CE
A
Tunnel
(full mesh)
VSI
PE3
PE4
 Full mesh of PW and tunnels deployed
 Tunnels
 Help transport the PW payload
 Aggregate traffic from multiple PWs
 Pseudowires – demultiplex the L2 traffic traversing tunnels
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
56
Metanoia, Inc.
Critical Systems Thinking™
Scaling VPLS: Hierarchical VPLS
 Base VPLS requires full mesh of VC LSPs between PE routers
 Adequate for PE routers in CO – multiple customers aggregated
 Inadequate for PE routers in MTU basements!
MTU
MTU
PE
PE
MTU
MTU
PE
PE
LSP explosion
Operational nightmare!
PE
MTU
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
57
Metanoia, Inc.
Critical Systems Thinking™
Hierarchical VPLS Advantages
MTU
MTU
PE
PE
Hub PE
MTU
PE
MTU
Core VC
LSP mesh
Spoke
VCs
(VLL or Q-in-Q)
PE
Benefits
 Simplifies signaling
PE
MTU
 Reduces pkt. replication
 Simplifies MTU
 Scalable inter-domain VPLS
 Simplifies new site addition
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
58
Metanoia, Inc.
Hierarchical VPLS: Case Study for
a Metro Region
Critical Systems Thinking™
100 MTUs; 10 customers/MTU; 2 VPLS/cust.; 100 stations/VPLS
VPLSs/MTU = 10x2 = 20
MTU100
CE
MACs/MTU = 20x100 = 2000
MTU1
PE
MTU 100
PE
MTU1
CE
MTU2
MTU99
MTU10
PE
PE
CE
MTU91
CE
Hub PE
MTU90
CE
PE
PE
MTU81
CE
PE
PE
MTU3
PE
MTU40
CE
MTU31
CE
MTU40
No hierarchy  PE supports
Hierarchy (10 MTU/PE)  PE supports
2000 MACs
2000 x 10 = 20,000 MACs
LDP/BGP sessions = (100x99)/2 x
20 = 245,000
LDP/BGP sessions = (10x9)/2 x 200 = 9000
Copyright 2007
All Rights Reserved
# of spoke VLLs = 10 x 20 = 200
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
59
Metanoia, Inc.
Critical Systems Thinking™
Benefits of IP/MPLS-based L2 VPNs
 Separation of administrative responsibilities
 Migration from traditional L2 VPNs: seamless transport of Ethernet
services
 Privacy of routing
 Layer 3 independence
 Less operational overhead
 Ease of configuration (?)
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
60
Metanoia, Inc.
Critical Systems Thinking™
Advanced Features:
Traffic Engineering,
Resilience, OAM, QoS
Metanoia, Inc.
Critical Systems Thinking™
Traffic Engineering Concepts
© Copyright 2006
All Rights Reserved
Metanoia, Inc.
Critical Systems Thinking™
Constraint Based Routing
 A class of routing systems that computes routes through a
network subject to a set of constraints and requirements
QoS-based Routing
 Path of flows determined by
 Knowledge of resource
availability in network
Policy-based Routing
 Path/routing decision based
on administrative policy
 QoS requirements of flows
 Can be on-line or off-line
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
63
Metanoia, Inc.
Critical Systems Thinking™
CB Routing System
 Inputs
Resources
 Flow/path attributes:
required b/w, hop count, ...
 Resource attributes:
Attributes
Topology
properties of nodes/links
 Network topology & state
Constraint-Based
Routing Process
 Outputs
 Computed feasible path
Feasible Path
ERO {1,3,4,5}
 Explicit route of the path
3
5
1
4
2
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
64
Metanoia, Inc.
Critical Systems Thinking™
MPLS-based Resilience for the Metro
© Copyright 2006
All Rights Reserved
Metanoia, Inc.
Critical Systems Thinking™
Fundamental Characteristics of RSVP
 Allows apps. to signal QoS requests to n/w, and n/w to respond
with success or failure
 Designed to transport
 Classification info. (Sender_Template)
 Allows flows with specific QoS reqs. to be recognized
 Traffic specs of source/sender (Tspec)
 QoS needs of receivers (Rspec)
 Soft-state protocol
 Path/Resv transmitted periodically to refresh reservation
 Refresh Reduction [RFC2961] has practically eliminated original
scalability concerns with use of soft state
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
66
Metanoia, Inc.
Critical Systems Thinking™
Basic Operation of RSVP-TE
Path (Label_Req)
A
Resv
Label=21
B
Path (Label_Req)
C
Resv
Label=49
Path Message
RSVP Header
D
E
Resv
Label=5
Resv
Label=7
Resv Message
SESSION
Application for which RSVP
reservation is to be made
SENDER_TEMPLATE
Identifies pkts. of the sender
RSVP Header
SESSION
STYLE
Specifies senders that may
use the reserved resources
LABEL
Label assigned to this hop
Record route taken by Path
SENDER_TSPEC
Defines traffic output by sender
LABEL_REQUEST
Request for label on this hop
RRO
ERO/RRO
Specific path to which flow is
to be bound
RSpec
SESSION_ATTRIBUTE
LSP attributes for this sender
SENDER_TEMPLATE
PHOP
Copyright 2007
All Rights Reserved
IP address of I/F that
transmitted Path Msg.
Same as that in Path Msg.
NHOP
QoS desired by receiver
Flow for which QoS is
desired
IP address of I/F originating
the Resv msg.
Flow Descriptor
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
67
Metanoia, Inc.
Fast Re-Route (FRR) using
RSVP-TE
Critical Systems Thinking™
 Rerouting is done when
 A better path is available
 Upon failure along LSP
Src
Originates LSPs
with IDs 1 and 2
Here they are treated as different
LSPs within the same Session
 Use SESSION Obj. & SE style
 Tunnel uniquely identified by
Rcvr
Tunnel ID in
Session Obj
LSP ID = L1
 Destination IP address
 Tunnel ID
 Ingress IP address
 Tunnel ingress made to appear
as 2 different senders to the
RSVP session (via LSP ID)
Copyright 2007
All Rights Reserved
LSP ID = L2
On these links the
LSPs share resources
LSPs 1 and 2 have a common SESSION Obj, but
a new LSP ID in the SENDER_TEMPLATE and a
different ERO (with possibly common hops)
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
68
TE with Constraint-based Routing
in a Nutshell
Operator Input
(Flow or LSP
Attributes)
Route Computation
Process
(on-line (CSPF) or offline)
TED
Demand or Traffic driven
LSP path selection
Critical Systems Thinking™
Enhanced IGP
Process
(OSPF-TE)
Network
Topology + State
Output
Computed
feasible path
(ERO)
Resource
Attributes
Metanoia, Inc.
Routing Table
(RIB)
Control driven route computation
and LSP path selection
Signaling Process
(RSVP-TE)
Link State
Database
(LSDB)
Standard IGP
Process (OSPF)
CONTROL PLANE
DATA PLANE
LSP
Establishment
MPLS LSPs
(Label Info. Base)
Copyright 2007
All Rights Reserved
Link Attribute
Modification
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
Forwarding
Info. Base (FIB)
69
Metanoia, Inc.
Critical Systems Thinking™
How it All Fits Together
Last-mile Ethernet
PBB clouds
CE3
LSP Tunnels
CE1
PE1
PE3
CE4
Pseudo-wires
PE2
IP/MPLS Core
CE2
Copyright 2007
All Rights Reserved
Attachment circuits
-- Physical (PDH/SDN)
-- Logical (FR, ATM, VLANs, tunnels)
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
70
Metanoia, Inc.
Critical Systems Thinking™
OAM: The Traditional Achilles Heel of
Ethernet
© Copyright 2006
All Rights Reserved
Metanoia, Inc.
Critical Systems Thinking™
Why Ethernet OAM?
 Current management protocols lack per-customer
granularity to handle Ethernet services
 Most management protocols operate are point-to-point
 Ethernet OAM can exploit multipoint capability
 Link management required for last-mile connection
 Similar to link mgt. in FR and ATM
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
72
Metanoia, Inc.
Critical Systems Thinking™
Ethernet OAM Types
 Service OAM
 e2e connectivity and fault mgt. per service instance
 Part of IEEE 802.1ag, CFM project
 Link OAM
 Monitoring & fault mgt of individual Ethernet link (physical/emulated)
 Part of IEEE 802.3, Clause 57 (formerly 802.3ah (not to be confused
with 802.1ah))
 Ethernet Local Mgt. Interface (E-LMI)
 Configuration & operational provisioning of customer edge device
 Part of MEF Standard MEF-16
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
73
Metanoia, Inc.
Critical Systems Thinking™
Service OAM
 Works on per-EVC basis
 Independent of underlying transport technology
 CFM messages
 Continuity Check Message
 Detects loss of service connectivity
 Link Trace Message
 Traces the path hop-by-hop (like IP traceroute)
 Loopback Message
 Detects whether target point is reachable (like ICMP Ping)
 AIS (Alarm Indication Signal) Message
 Asynchronous notification to indicate fault
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
74
Metanoia, Inc.
Critical Systems Thinking™
Link OAM
 Discovery
 Identifies devices at both ends of the link
 Link Monitoring
 Detects link faults
 Statistics of packet errors
 Remote Failure Indication
 Conveys loss-of-signal indication to peers, due to poor SNR, power
failure, or other critical events
 Remote Loopback
 Determines quality of link during installation and troubleshooting
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
75
Metanoia, Inc.
Critical Systems Thinking™
E-LMI
 Provides local configuration & operational parameters to
customer edge
 VLAN-EVC mapping
 QoS profiles of EVC
 Reduces configuration errors, improves performance
 Dynamic EVC management
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
76
Metanoia, Inc.
Critical Systems Thinking™
Quality-of-Service: Ah! that elusive QoS
© Copyright 2006
All Rights Reserved
MPLS and Quality-of-Service for
Ethernet Services
Metanoia, Inc.
Critical Systems Thinking™
 MPLS supports (not extends) a packet-based QoS model
 MPLS does not run in hosts (only in metro/core routers)
 QoS, however, is an end-to-end mechanism
 MPLS helps carriers offer QoS-enabled services efficiently
 Can support MEF QoS model via DiffServ QoS framework
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
78
Metanoia, Inc.
Critical Systems Thinking™
Differentiated Services Framework
 Traffic flows aggregated into small # of classes
Drop Precedence
 Per-flow state is not required
 More scalable than IntServ
3
 Class encoded in IP header via
DiffServ Code Point (DSCP)
 Edge router …
 Classifies packets to DifServ classes
2
1
Class Priority
EF
AF1x
DSCP
101110
001xx0
AF2x
01xx10
AF3x
11xx10
AF4x
1xxx10
BE

Best Effort (BE)

Expedited Forwarding (EF)
 Minimal delay & loss

Assured Forwarding (AF)
 4 classes
 3 drop precedence’s each
 DSCP identifies Per Hop Behavior (PHB)
Copyright 2007
All Rights Reserved
 12 possibilities total
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
79
Metanoia, Inc.
Critical Systems Thinking™
Differentiated Services Architecture
Diffserv Domain
Core Functions
Edge Functions
EF
Traffic Conditioning
Meter
Colored packet
(marked DSCP)
Strict
Priority
AF
Classifier
Marker
Shaper
Aggregate
PHBs
Scheduling
BE
WFQ
Queueing
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
80
MPLS Support of DiffServ:
Mapping DSCPs to LSPs (or labels)
Metanoia, Inc.
Critical Systems Thinking™
 Map DSCP  EXP bits in MPLS “shim” header
 6 DS bits (64 PHBs) and only 3 EXP bits (8 classes)!
 Complete mapping is infeasible
 For many practical cases, 8 PHBs may suffice
IP Header
MPLS “shim” header
6 bits
DSCP
DSCP
DS byte
Label
EXP
S
TTL
3 bits
Results in an LSP called an E-LSP
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
81
MPLS Support of DiffServ:
Mapping DSCPs to LSPs (or labels)
Metanoia, Inc.
Critical Systems Thinking™
 Map {PHB, FEC}  MPLS Label
 That is, provide the info. in the label itself!
 Requires enhancing the label distribution protocols
 Use EXP bits for drop precedence
 That is to determine different PHBs of a PHB scheduling class
6 bits
DS class drop
precedence
DS class: EF, AFx
DSCP
DSCP
Label
EXP
S
TTL
3 bits
DS byte
IP Header
MPLS “shim” header
Results in an LSP called an L-LSP
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
82
Metanoia, Inc.
Critical Systems Thinking™
Conclusions and Discussion
Metanoia, Inc.
Critical Systems Thinking™
Conclusions
 Ethernet poised to be dominant choice in metro networks
 Reduces capex and opex for providers
 Enables new revenue generating services
 802.1ad provider bridge with OAM of 802.1ag …
 … a choice at the edge
 Two architectures emerging for Ethernet in the metro core
 Provider Backbone Transport (PBT)
 IP/MPLS-based L2 VPNs
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
84
Metanoia, Inc.
Critical Systems Thinking™
Thank You!
Questions?
Metanoia, Inc.
Critical Systems Thinking™
Glossary
AC
Attachment Circuit
DS
DiffServ
ACL
Access Control List
DSCP
DiffServ Code Point
AF
Assured Forwarding
EF
Expedited Forwarding
API
Application Programming Interface
E-LMI
Ethernet-Local Management Interface
AS
Autonomous System
E-LSP
EXP mapped LSP
ATM
Asynchronous Transfer Mode
EPL
Ethernet Private Line
BA
Behavior Aggregate
ERO
Explicit Route Object
B-DA
Backbone Destination Address
E-UNI
Ethernet UNI
B-DA
Backbone Source Address
EVC
Ethernet Virtual Circuit
BE
Best Effort
EVPL
Ethernet Virtual Private Line
B-FCS
Backbone Frame Check Sequence
EXP
Experimental (EXP bits in MPLS "shim"
header)
EXP
Experimental Bits
FCS
Frame Check Sequence
FEC
Forwarding Equivalence Class
FIB
Forwarding Information Base
FR
Frame Relay
GR
Graceful Restart
H-QoS
Hierarchical Quality-of-Service
H-VPLS
Hierarchical VPLS
IPTV
IP Television
BGP
Border Gateway Protocol
CBS
Committed Burst Size
CE
Customer Edge (router)
CES
Core Ethernet Switch/Bridge
CFM
CIR
Committed Information Rate
CO
Central Office
DA
Destination Address
DS
DiffServ
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
86
Metanoia, Inc.
Critical Systems Thinking™
Glossary
L2
Layer 2 (Data Link Layer; MAC Layer)
OSPF
Open Shortest Path First
L3
Layer 3 (Network or IP Layer)
P
Provider (router)
LAN
Local Area Network
PB
Provider Bridging
LDP
Label Distribution Protocol
PBB
Provider Backbone Bridging
LER
Label Edge Router
PBT
Provider Backbone Transport
LIB
Label Information Base
PDH
Pleisosynchronous Digital Hierarchy
L-LSP
Label inferred LSP
PE
Provider Edge (router)
LSP
Label Switched Path
PHB
Per Hop Behavior
LSR
Label Switching Router
PIR
Peak Information Rate
MAC
Medium Access Control
PSN
Packet Switching Network
MBS
Maximum Burst Size
P-VLAN
Provider VLAN
MEF
Metro Ethernet Forum
PW
Pseudo-Wire
MEN
Metro Ethernet Architecture
QoS
Quality-of-Service
MPLS
Multi-Protocol Label Switching
RIB
Routing Information Base
MSTP
Multiple Shortest Path Tree
RSTP
Rapid Spanning Tree Protocol
MTU
Multi-Tenant Unit
NG
Next Generation
RSVP-TE
Resource Reservation Protocol - Traffic
Engineering (RSVP protocol with MPLS
traffic engineering extensions)
NGN
Next-Generation Network
SA
Source Address
NNI
Network Network Interface
SDH
Synchronous Digital Hierarchy
OAM
Operations, Administration, and Management
SONET
Synchronous Optical Network
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
87
Metanoia, Inc.
Critical Systems Thinking™
Glossary
SPT
Shortest Path Tree
VRF
Virtual Routing and Forwarding
ST
Spanning Tree Protocol
VSI
Virtual Switching Instance
STP
Spanning Tree Protocol
WFQ
Weighted Fair Queuing
TDM
Time-Division Multiplexing
TE
Traffic Engineering
TM
Traffic Management
TTL
Time to Live
UNI
User Network Interface
VCI
Virtual Circuit Identifier
VFI
Virtual Forwarding Instance
VID
VLAN Identifier
VLAN
Virtual LAN
VLAN
Virtual LAN
VOQ
Virtual Output Queue
VPI
Virtual Path Identifier
VPLS
Virtual Private LAN Service
VPN
Virtual Private Network
VPWS
Virtual Private Wire Service
VR
Virtual Router
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
88
Metanoia, Inc.
Critical Systems Thinking™
Readings and References (1)
 MEF 4: Metro Ethernet Network Architecture Framework Part 1 Generic
Framework
 MEF 6: Metro Ethernet Services Definition Phase 1
 MEF 10.1: Metro Ethernet Services Attributes Phase 2
 MEF 16: Ethernet Local Management Interface
 IEEE 802.1d/q WG: “Media Access Control (MAC) Bridges,” IEEE 1998
 IEEE 802.1s, “Multiple Spanning Tree,” IEEE 2002
 IEEE 802.1ah, “Provider Backbone Bridges,” Work in Progress
 Documents on the MEF and IEEE 802.1 and 802.3 WG web sites
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
89
Metanoia, Inc.
Critical Systems Thinking™
Readings and References (2)

L. Andersson and E. Rosen, “Framework for Layer 2 Virtual Private
Networks (L2VPNs),” RFC 4664, September 2006

K. Kompella and Y. Rekhter, Eds., “Virtual Private LAN Service: Using
BGP for Autodiscovery and Signaling,” RFC 4761, January 2007

V. Kompella and M. Lasserre, Eds., “Virtual Private LAN Service: Using
Label Distribution Protocol for Signaling,” RFC 4762, January 2007

S. Bryant and P. Pate, Eds. “Pseudo Wire Emulation Edge-to-Edge (PWE3)
Architecture,” RFC 3985, March 2005

L. Martini et al, Eds., “Pseudowire Setup and Maintenance Using the Label
Distribution Protocol (LDP),” RFC 4447, April 2006

Documents on the L2 VPN, PWE3, MPLS, and CCAMP WG’s of the IETF
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
90
Metanoia, Inc.
Critical Systems Thinking™
Additional Slides
Label Assignment and Distribution
(control component)
Data
Labels
Metanoia, Inc.
Critical Systems Thinking™
Data
Labels
Downstream
Upstream
Ordered
Solicited (On Demand)
Unsolicited
Solicited
Unsolicited
Independent
Solicited (On Demand)
Unsolicited
Solicited
Unsolicited
Direction from which labels flow
Whether LSR waits to hear from
its upstream/downstream nbrs.
before responding to a request
for label(s)
Refers to whether LSR distributes
labels on demand or voluntarily
Label Retention: Liberal or Conservative
Whether LSR keeps labels from a neighbor
who is not currently the next hop for a FEC
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
92
Metanoia, Inc.
Critical Systems Thinking™
A Word on Reservation Styles
S1
 Always chosen by the receiver
Unique label/sender
Distinct reservation
per sender
 Two styles apply with RSVP-TE
S2
 Fixed Filter (FF)
 Distinct reservation for traffic
Link (i,j)
from each sender
 Needs unique label per sender
S1
S3
Common reservation
shared by all senders
 Shared Explicit (SE)
 Common resvn. for traffic from
the senders specified by rcvr.
 May assign unique label/sender
S2
 Useful for p2p or mp2p LSPs
Link (i,j)
Different senders may
have different labels
Copyright 2007
All Rights Reserved
S3
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
93
Metanoia, Inc.
Critical Systems Thinking™
LDP versus BGP Signaling
PE
PE
PE
PE
Targeted
LDP
i-BGP
PE
PE
PE
Targeted LDP
PE
RR
PE
PE
BGP-based Signaling
 LDP session full mesh b/ween PE’s
 RR’s reduce full mesh to 2 sessions/PE
 PE’s exchange labels directly
 New PE  reconfig. mesh at all PE’s
 Cannot direct label mapping to a
specific peer  need label ranges
 FIB per VPLS per PE
 New PE  peering session only w/ RRs
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
94
Metanoia, Inc.
Critical Systems Thinking™
L2 VPNS with BGP
 Autodiscovery + signaling, together via BGP with RTs (per slide 74)
 PE configured with its VPLS ID (if VPLS)
 Transmits VPLD ID or identity of attached CE’s to peer PE’s
 Includes demux value for each BGP NLRI (as a label range)
 Selection algorithm allows each remote PE to pick correct label for
sending traffic to advertising PE
BGP NLRI for VPLS
BGP NLRI for L2 VPN
Length (2 octets)
Length (2 octets)
RD (8 octets)
RD (8 octets)
VE ID (2 octets)
CE ID (2 octets)
VE Block Offset (2 octets)
Label blk offset (2 octets)
VE Block size (2 octets)
Label Base (3 octets)
Label Base (3 octets)
Circuit Status Vector
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
95
Metanoia, Inc.
Critical Systems Thinking™
BGP-based L2 VPN (VPWS)
DLCI=[11,12,…, 30]
Label block offset=0
Label base = 3000
Label range = 20
CE3
DLCI=[101, 102, …, 120]
11
CE1
103
12
CE4
1003
Label block offset=0
Label base = 1000
Label range = 20
PE1
3001
PE3
2003
PE2
Label block offset=0
Label base = 2000
Label range = 20
CE2
3002
IP/MPLS
Core
403
DLCI=[401, 402, …, 420]
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
96
Metanoia, Inc.
Critical Systems Thinking™
BGP-based L2 VPN (VPLS)
CE3
CE1
CE4
3001
PE1
PE3
3002
PE2
Label block offset=0
Label block size = 10
Label base = 3000
VE ID = 3
IP/MPLS
Core
CE2
Copyright 2007
All Rights Reserved
Next-Generation Systems & Networks Workshop, 17th July. 2007, Bangalore, India
97