A Practical Approach for Providing QoS: MPLS and DiffServ

1
A Practical Approach for Providing QoS MPLS and
DiffServ
MPLSMaximizing the Performance and
Profitability of Optical and Data Networks July
12 - 13, 2001 Dublin, Ireland

• Thomas Telkamp
• Director Data Architecture and Technology
• Global Crossing Telecommunications, Inc.
• telkamp_at_gblx.net

2
Agenda

• Global Crossing MPLS deployment
• Quality of Service?
• A Practical Approach
• Network Design and Capacity Planning
• Differentiated Services
• MPLS for Traffic Engineering
• Fast Reroute and Per-Class TE
• Queuing and Scheduling
• Conclusion

3
Global Crossing IP Backbone Network
PEC
AC-2
AC-1
North American Crossing
PC-1
MAC
EAC
PAC
SAC
Based on announced network
4
GBLX MPLS Deployment

• Operational since 2Q 1999
• Traffic Engineering
• IP TTL issues
• Worldwide MPLS mesh 1Q 2001
• Currently over 6000 LSPs
• Network
• Cisco and Juniper routers
• OC-48 wavelengths
• Covering Asia, US, South America and Europe
• New Services VPN (L2/L3)

5
MPLS Traffic Engineering
6
(No Transcript)
7
Quality of Service?Based on a paper with XiPeng
Xiao (Photoris, Inc.) and Lionel M. Ni (Michigan
State University)

• Best Effort (e.g. Internet)
• Real-time/Mission-critical traffic (e.g. Voice)
• Increase revenue by value-added services
• Two extremes
• Overprovisioning of bandwidth without additional
  mechanisms
• Sophisticated mechanisms such as per-flow
  classification/policing/queuing and scheduling

8
What Causes Problems?

• Overloaded servers, or access to servers
• Web, E-mail, etc.
• TCP stack implementations
• Link failures
• fiber cuts
• transmissions equipment failures
• Router failures
• complex software
• early deployment of features
• configuration

9
A Practical Approach

• Good Network Design
• Differentiated Services (DiffServ)
• Traffic Engineering
• Traffic Protection (Fast Reroute)
• Class-based Queuing
• Not
• Extremely complex schemes (e.g. per-flow)
• affecting equipment reliability
• difficult to configure and manage

10
Network Design

• Avoid single points of failure
• No bottlenecks in normal condition
• Overprovisioning
• with use of TE network can handle all traffic,
  even when the most critical links fails
• Routing (IGP and BGP)
• Security and Denial of Service attacks
• Capacity Planning

11
Differentiated Services

• How many classes?
• What are the targeted applications for each
  class?
• Can end users distinguish between classes?
• Example
• Class 1 Real-time
• application voice
• Class 2 Assured
• application trading, non-interactive audio and
  video
• Class 3 Best Effort
• application Internet

12
MPLS Deployment

• Traffic Engineering
• Avoid congestion caused by uneven traffic
  distribution
• Macro control
• Constraint based LSP setup
• Two LSP meshes
• Real-time traffic vs Assured/Best Effort
• Classification based on interface or multi-field
  lookup
• Different metrics
• LSP Hierarchy
• Scalability and VPNs

13
MPLS LSP Deployment
14
Traffic Protection

• IGP convergence (OSPF/IS-IS) takes seconds
• But can be improved by timer and SPF tuning
• see Packet Design paper
• MPLS Fast Reroute
• Link or Node protection
• Pre-configured patch LSPs (sub-optimal)
• Use for real-time traffic only, or for all
  traffic (based on implementation)

15
MPLS Fast Reroute

• Protecting router switches traffic to
  pre-configured patch LSP after failure detection
  (fast)
• Ingress router reroutes LSP (slow)

16
Per-Class Traffic Engineering

• Avoid concentration of real-time traffic at any
  link
• Set upper limit on bandwidth reservations per
  class
• E.g. max. 40 of a link for VoIP traffic
• IETF Internet Draft(s) on Diff-Serv-aware MPLS
  Traffic Engineering (Francois Le Faucheur, et
  al.)

17
Class-based Queuing

• Prefer higher classes during congestion
• sub-optimal fast-reroute period
• major failures
• Different queuing/scheduling mechanisms
• Strict Priority Queuing
• Jitter control for EF traffic
• WRR/WFQ
• and combinations
• Configuration issues...

18
Random Early Detection

• Buffer Management
• prevent tail-drop
• TCP oscillations and synchronization
• RED drops based on average queue length
• WRED drops with different probability for each
  class
• Only during congestion
• Not used to guarantee bandwidth!

19
Conclusion

• Use combination of good network design,
  over-provisioning and MPLS/DiffServ
• Use Traffic Engineering to prevent congestion
• Use fast reroute and priority queuing for
  real-time traffic
• Use WRR/WFQ to differentiate between Assured and
  Best Effort traffic
• Too complex and too many features will make the
  network unreliable/unstable

20
Questions?