Transcript Experiences with Deploying a Global IP/MPLS Network

China International Summit
Technology Convergence &
Next Generation Networks
September 25 - 26, 2001
Beijing, P.R. China
Experiences with Deploying a
Global IP/MPLS Network
Thomas Telkamp
Director Data Architecture and Technology
Global Crossing Telecommunications, Inc.
[email protected]
Agenda
• Global Crossing Network
• IP Network Evolution and Philosophy
• MPLS Deployment
• IP VPNs
• Multi-Service and DiffServ
• New Features
• Other Issues
• Summary
Global Crossing History
• The First Independent Global Fiber Network
 Started operations in March 1997
 First segment in service on May 26, 1998
• Expanding Network and Services by acquisitions:
 Frontier Telecommunications, Sept 1999
 Racal Telecom, Nov 1999
 Hutchison Global Crossing, Jan 2000
 IXNET/IPC, June 2000
• Global Network
 100,000 route mile, 27 countries, 200 major cities
Cable ship...
…and the cable.
Global IP Backbone Network
Global IP Network
• OC-48c/STM-16c (2.5Gbps) IP backbone
 Some 10Gbps segments operational (e.g. Atlantic)
• Multi-Vendor:
 Cisco GSR 12000/12400
 Juniper M20/M40/M160
• BGP and IS-IS routing protocols
• Internet Access & Transit Services
• IP VPN, Layer 3 and Layer 2
• Global MPLS and DiffServ deployment
IP Network Evolution
• Network technology and usage changes over time
• Roadmap
 US domestic backbone for Internet traffic
 MPLS deployment for Traffic Engineering
 Global extension of the network
 RFC2547 VPNs
 DiffServ deployment for multi-service
 Optimization for critical services (e.g. voice)
Network Philosophy
• No bottlenecks in normal condition
• Overprovisioning
 with use of Traffic Engineering network can handle
all traffic, even when the most critical links fails
• MPLS Traffic Engineering to prevent congestion
• DiffServ to manage congestion
• Too complex and too many features will make the
network unreliable/unstable
MPLS Deployment
• Operational since 2Q 1999
 Traffic Engineering
 IP TTL issues
• Worldwide MPLS mesh 1Q 2001
• Currently over 6000 MPLS LSPs
• Support for RFC2547 VPNs
 MPLS/BGP technology
MPLS LSP Deployment
MPLS Deployment Experiences and
Challenges
• Router vendor interoperability
 Cisco and Juniper
 New features...
• Requires seamless network
 Single AS
 Single IGP (no multiple areas/levels)
• NOC education and training
• Network Management
 Software to support Traffic Engineering
Why Traffic Engineering?
MPLS Traffic Engineering
IP VPNs
• RFC2547 VPN
 Based on existing MPLS and BGP protocols
• Minimal impact on operations if already running a
MPLS based Internet backbone
• Layer 2 variant under development
DiffServ Deployment
• Increase revenue by value-added services
 Best-Effort
• Internet
 Assured
• trading and non-interactive audio and video)
 Real-time
• voice
• Prefer ‘higher’ classes during congestion
 major failures
 guarantee delay and jitter
DiffServ Challenges
• How many classes?
 What are the targeted applications for each class?
 Can end users distinguish between classes?
• How to implement these classes?
 Different queuing/scheduling mechanisms
• Strict Priority Queuing
• WRR/WFQ
• and combinations
 Configuration and Monitoring issues...
New Features
• MPLS Fast Reroute
 Provides SDH like restoration times for critical
services (e.g. voice and trading)
• Per-Class Traffic Engineering
 Avoid concentration of real-time traffic at any link
 Set upper limit on bandwidth reservations per class
• IGP tuning for better performance
• IS-IS parameters and configuration
Other Issues...
• What about GMPLS?
 What are the benefits to the IP layer?
 What problems does it solve?
• Do we still need ATM?
 Cell based networks have different characteristics
than packet based networks
 CBR service for Circuit Emulation
• Should we build ‘Pure IP’ networks?
 There are alternatives to MPLS Traffic Engineering
and MPLS based VPNs
Summary
• Evolve existing network or build new network
based on applications and requirements:
 MPLS for Traffic Engineering
 DiffServ for service differentiation
 Advanced features for critical applications
• Don’t underestimate the operational aspects of
new technology in the network
• Too complex and too many features will make the
network unreliable and unstable
Questions?