Transcript Using Multi-Layer Routing to Provision Services across

Using Multi-Layer Routing to
Provision Services across
MPLS/GMPLS Domain
Boundaries
Andrew G. Malis
Chief Technologist, Tellabs
Chairman and President, MFA Forum
1
Discussion of Convergence is
Everywhere
 “…the access infrastructure telcos are building to support IPTV
services will finally give them the ability to converge voice,
video and data onto a single network…”
– Americas Network
 Pseudo-Wires are “the solution for convergence in future
telecom networks, because it preserves profitable legacy
services even as it enables the creation of a truly nextgeneration network” – Heavy Reading
 “3GPP defines a new subsystem to enable the convergence of
voice and data applications and the harmonization of various
mobile network technologies over IP”
– Deutsche Bank
2
Convergence is changing
the face of the network
The Benefits of Convergence
 Services reduced to applications on converged infrastructure
 Reduced operations
 Reduced core cap-ex
 New services can be provided
 Individual Networks per Service limited service interaction
3
The details in getting there are not small…
The Reality of Convergence
 Non-native service delivery
 End User Services are always layered on something else
 Which Technology for Convergence?
 WDM
 Layer 1 (GMPLS)
 Layer 2/3 (MPLS)
 The whole network will not be converged overnight
 Too many existing services already deployed
 The whole network will never be completely converged
 Fully depreciated equipment still generates revenue
4
How can this all be supported
operationally?
Access A
(WDM)
Islands of
Technology
CPE-A
Core
(MPLS Pseudo Wire)
CPE-B
5
Access B
(SONET/SDH)
Access A
(WDM)
Different
Organizations
Manage These
Technologies
CPE-A
Core
(MPLS Pseudo Wire)
CPE-B
6
Access B
(SONET/SDH)
We will fill
your order
within 3
months
Access A
(WDM)
Provisioning
Service is
expensive and
time consuming
CPE-A
Core
(MPLS Pseudo Wire)
CPE-B
I forgot that
we ordered
this service!
7
Access B
(SONET/SDH)
Access A
(WDM)
OCN
UNI
WDM Network
GE/NNI
T1, T3, OCN
UNI
What would be
best?
GE/NNI
OCN
UNI
GE/NNI
GE/NNI
CPE-A
T1, T3, OCN
UNI
10/100
BPON
BPON
GE/NNI
BPON
GE/UNI
Core
(MPLS Pseudo Wire)
10/100
GE/MPLS
CPE-B
Access B
(SONET/SDH)
OCN/NNI
10/100
nxT1
T1, T3, OCN
UNI or Null
T1, T3, OCN
UNI or Null
OCN/NNI
OCN/NNI
8
MPLS Network
OCN/NNI
nxT1
10/100
GE/UNI
OCN/NNI
STS1 Network*
OCN/NNI
Sure, we’re
setting your
service up
as we speak
OCN
UNI
WDM Network
GE/NNI
T1, T3, OCN
UNI
GE/NNI
OCN
UNI
GE/NNI
GE/NNI
T1, T3, OCN
UNI
10/100
BPON
BPON
GE/NNI
BPON
GE/UNI
10/100
End-to-end Signaled
Service Provisioning
GE/MPLS
10/100
GE/UNI
nxT1
MPLS Network
OCN/NNI
OCN/NNI
nxT1
10/100
T1, T3, OCN
UNI or Null
T1, T3, OCN
UNI or Null
Wow,
that was
fast!
9
OCN/NNI
OCN/NNI
OCN/NNI
STS1 Network
OCN/NNI
How Can This Be Accomplished?
 Common flexible control method that understands layering
 Path Computation that understand layered networks
 Provides service routing given view of potential/available server
layer resources
 Signaling mechanisms that coordinate calls in different
layers
 OSSes that can handle integrated views of layer networks
 Relate services requests (client layer calls) to server resources
in use
 Definitions for server layer resources other than SONET/SDH
 An Added Benefit: Integrated operations
 Operations convergence possible due to Common control
methods
10
Why Use GMPLS?
 Establishes a common control plane for different networking
technologies
 Converge Packet, Cell, TDM, and Optical administrative controls
 Automates connection management for all traffic types
 Path setup and management (for Packet, Cell, TDM, and Optics)
 Handles topology changes automatically
 Self-discovery and dynamic configuration of network resources
 Provides static and dynamic path reroutes and restoration
 Supports Peer-to-Peer and Overlay network models
 Integration of optical switches, optical transport, and label
switching routers
11
Making Path Computation Layer
Aware
 Routing today treats different technologies as separate
topology graphs
SONET Network
PON Access
MPLS-TE
Network
12
WDM Network
Making Path Computation Layer
Aware
 Normal Path Computation cannot find paths between
endpoints on different islands
SONET Network
PON Access
MPLS-TE
Network
13
WDM Network
Making Path Computation Layer
Aware
 As a result, end-to-end services are separately routed by
each island, and interconnected by hand
SONET Network
PON Access
MPLS-TE
Network
14
WDM Network
Making Path Computation Layer
Aware
 By merging the graphs, adding client-layer matrixes and
client/server adaptation, the graph becomes continuous
SONET Network
PON Access
MPLS-TE
Network
15
WDM Network
Making Path Computation Layer
Aware
 GMPLS routing extensions
 Link Attribute Announcements that:
 remove ambiguity of adaptations supported
 announce adaptations in a technology independent manner
 necessary to allow for source routing to be done anywhere
 include link costs that take into account:
 different costs for each layer supported by a link
 cost to utilize adaptation
 Path Computation Algorithm that:
 understands multiple matrices per node
 updates “signal stack” when adaptations are pushed/popped
16
IP & Optical Layer Combination
Practical Example
PE
MPLS Layer
P
GMPLS Layer
OXC
Major Objectives:
-Eliminate electrical packet processing in core
-Minimal or no change to existing MPLS routers
-Take advantage of TE in both networks – carry out multi-level TE
17
MPLSGMPLS Interworking:
Overlay vs. Peer to Peer
Overlay Model
 Optical Domain is not visible to IP domain
 MPLS domain can not perform efficient TE
 Do not have to update MPLS nodes to
GMPLS
Peer to Peer Model
 Optical Domain is visible to IP domain
 Optimal route assured since the MPLS
network can perform efficient TE by
understanding GMPLS network resources
 Must update MPLS node to GMPLS
Can we take advantage of both models?
18
Optimal Design of an MPLS Core Network
P routers
Service
Demands
PE routers
 LSP mesh between PEs originally traverses P routers
 Direct PE-PE tunnels preferable where sufficient
commonality of traffic endpoints exists
 Move LSPs to direct route reducing P-router load
19
Feasible and cost-effective with GMPLS
Essence of Multi-Layer Core Optimization
PE
LSPs
Original
Direct
MPLS
Layer
P
Transport
Layer
 Original tunnel




20
Six router interfaces
 Direct tunnel
Forwarding on two P routers
Ten transport interfaces
 Two router interfaces
 Six transport interfaces
2x bandwidth on links to P routers
GMPLS affords dynamic, optimal size direct tunnels –
without necessity to wait for transport service orders
Standards Activities
 GMPLS
 IETF – Multi-region Network
 draft-shiomoto-ccamp-gmpls-mrn-reqs-03.txt
 draft-leroux-ccamp-gmpls-mrn-eval-02.txt
 ASON
 ITU – Multi-layer calls and Multi-layer Routing
 G.8080 Amendment 2
 Q12/15 March 2005 Interim WD53r1
 OIF – UNI 2.0 Ethernet Services & E-NNI Routing
 Multi-layer call support
 Architecture supports multi-layer routing (oif2004.477)
 Demonstrated at Supercomm 2005
21
Conclusion
 Convergence is occurring in the network
 New Services
 Lower CapEx and OpEx for network operators
 MPLS has become the industry-standard mechanism for
data network convergence (Layer 3 and Layer 2 services)
 GMPLS is being deployed as optical network equipment
replacement picks up
 Flexible control methods that support layering are in
progress
 Standardization activities underway in ITU, IETF and OIF
 Multi-vendor interoperability already demonstrated
22