Two-layer Restoration Scheme for IP over Optical Networks with MPLS

1. 8th IEEE International Conference on Communications Systems Two-layer Restoration Scheme for IP over Optical Networks with MPLS Jia Ke, L. Mason, Q. Yang ICIS, School of EEE, Nanyang Technological University jiake2000@hotmail.com

2. IP-over-WDM Networks • WDM is a frequency based multiplexing technique to increase the bit rate. Each fiber has a set of parallel optical channels, which use slightly different light wavelengths. • An unprecedented shift has occurred in traffic pattern from fixed, configured, connection-oriented services (e.g. voice service) to dynamic, connectionless IP services.

3. IP-over-WDM Networks (cont) • Optical networking evolves towards an integrated IP-over-WDM network architecture.

4. Multi-protocol Label Switching • In MPLS, packets are forwarded based on the appended labels. MPLS separates the routing decisions and forwarding of the data. • Connection-oriented label switched path (LSP) is set up between label switching routers (LSR) for connectionless IP packet transfer. • MPLS separates the routing decisions and forwarding of the data.

5. Multi-protocol Lambda Switching • Porting MPLS to the photonic domain results in the Multi-protocol Lambda Switching (MPλS). • Each wavelength in the fiber is treated as a label. • A Lightpath which is equivalent to an LSP in MPLS is setup between Optical Cross Connect (OXC).

6. Survivability of Networks • Two types of fault-management: protection and restoration • Protection: the spare capacity is reserved during working path setup. Costs more resource, static, pre-planned. • Restoration: the spare capacity is available after the fault occurrence. Dynamic, based on updated topology.

7. Path switching and link switching • Both have two techniques: path switching and link switching • Path switch: the failure is addressed at the path endpoints. • Link switch: the failure is recovered around failed links

8. MPLS recovery Mechanisms • Path Protection • a pre-established physically disjoint backup LSP is set up spanning the working LSP from ingress LSR to egress LSR. Figure Shows the path protection, the working LSP is LSR1→LSR3→LSR5→LSR7→LSR9, and the backup LSP is LSR1→LSR2→LSR4→LSR6→ LSR8→LSR9. They span the same protection ingress LSR LSR1 and protection egress LSR LSR9. When the PIL LSR1 gets the fault notification message, it will switch the traffic from the working LSP to the backup LSP.

9. MPLS recovery Mechanisms (cont) • Local Protection • protection is implemented on a link switching base. At the time that a failure occurs, the PSL LSR 5 immediately switches traffic through the backup LSP LSR5→LSR6→LSR8→LSR7 to the PML LSR7. At LSR7 the working LSP and the backup LSP will merge into one outgoing LSP.

10. MPLS recovery Mechanisms (cont) • Local loop-back Mechanism • A hybrid scheme combines the best characteristics of both path and local protection schemes. • A backup LSP is provided in the opposite direction and concatenated to a physically disjoint LSP.

11. MPLS recovery Mechanisms (cont) • Rerouting • A restoration mechanism, based on the real-time establishment of the backup LSP. • Long recovery time due to the routing table updating. • Some improvement is available e.g. explicit failure notification and Fast Topology-driven Constraint-based Rerouting (FTCR).

12. MPLS recovery Mechanisms (cont) In this example, after the failure occurs, the LSR1 finds the route of the working LSP (LSP1: LSR1→ LSR3→LSR5 →LSR7→ LSR9) needs to modify from the resluts of routing table update or explicit failure notification. Thus, LSR1 dynamically re-calculate the working LSP and establish the new working LSP: LSP1’ while tear down the previous LSP1.

13. Recovery in the Optical Layer • It is possible to perform protection in the optical layer while restricted to the physical characteristics of optical networks. • A complete wavelength is consumed by the backup lightpath. • Wavelength continuity constraint will be taken into consideration. • Restoration in the Optical layer • Does not suffer so much from the high capacity cost problem as does the protection mechanism. • Also should consider the wavelength continuity constraint.

14. Single-layer Survivability • Optical layer provides resilience • Recovery actions are performed on the coarsest granularity. A single element failure is treated and fewer recovery actions are taken. • Failures affecting higher layer, e.g.. Isolated LSR due to the failure of the OXC can not be fixed by optical layer. • IP layer provides resilience • Many recovery actions are needed, due to the finer granularity of the LSPs. • Faces the complex secondary failures in the virtual topology due to a single element failure in the optical layer. • allows the differentiation between individual LSPs, based on their service class with different reliability.

15. Our proposal: a novel joint two-layer recovery scheme • Recovery actions are taken both inoptical and IP layer. • Recovery starts from the optical layer and the IP/MPLS layer is activated if the optical layer cannot restore all affected traffic. • The spectrum of recovery mechanisms can be deployed in either layer. Here, path rerouting is used in both layers. • Some inter-working mechanisms exist for handing over the recovery actions.

16. Our proposal: a framework

17. Our proposal: advantages • Advantages • High recovery speed due to recovery at optical lightpath layer, close to the failure, batch recovery.It leads to less traffic loss and high throughput. • A finer granularity restoration will applied for some traffic flows which can not be recovery by coarse restoration at optical layer.

18. Our proposal: influence on network parameters • Influence of Traffic Demand • spare wavelength coefficient θ denotes the ratio of spare wavelengths to all available wavelengths. • With larger θ, more affect traffic is recovered in the optical layer with a very fast restoration speed.

19. Our proposal: influence on network parameters (cont) • Influence of granularity • Granularity coefficient ηrepresents the number of LSPs in each lightpath. • the MPLS layer restoration speed for larger η is slower than it is for smaller η. A larger η (finer granularity) leads to a little higher maximum recovery ratio.

20. Our proposal: influence of inter-working mechanisms • Hand-off timer vs. Recovery token • recovery with the hold-off timer starts about 150 ms later than the recovery using a recovery token. It degrades the recovery performance. • The hold-off timer is easier to implement compares with the recovery token mechanism.

21. Our proposal: influence of wavelength conversion • Full conversion vs. non-conversion • In current optical network, it is not economical to place wavelength converters everywhere. • The wavelength continuity constraint significantly deteriorates the recovery performance. • It only has a minor influence on the IP/MPLS layer recovery, since the intermediate LSRs perform the O-E-O conversion.

22. Summary • Recovery in IP/MPLS layer. • Recovery in MPλS layer. • A combination: an escalation two-layer recovery scheme.

23. Thank you!