Routing and Wavelength Assignment in All-Optical DWDM Transport Networks with Sparse Wavelength Conversion Capabilities Ala I. Al-Fuqaha, Ph.D.
Overview Transport Network Architectures: Current Vs. IP over DWDM Introduce Routing and Wavelength Assignment (RWA) problem in networks with Sparse wavelength conversion (RWA-SWC) Propose Integer Linear Programming (ILP) formulation for RWA-SWC Propose OSPF Extension (new LSAs) in support of all-optical DWDM networks with sparse wavelength conversion Propose new OSPF LSA origination Policies (Immediate/Fuzzy) 2
Will not talk about 3 Propose new Most-Contiguous Wavelength Assignment Heuristic Propose new fuzzy-based approach for route selection
Part 1 4 Today s Transport Network - Architecture IP Routers Voice Switches Services Overlay DS-3 W-DCS W-DCS Overlay W-DCS W-DCS OC-3, OC-12 DS-3, OC-3, OC-12 ATM B-DCS B-DCS Overlay ATM ATM B-DCS B-DCS OC-12, OC-48 DS-3, OC-3, OC-12, OC-48 ADM Sonet ADM Overlay ADM ADM
Today s Transport Network Architecture - Disadvantages Part 1 5 Functional Overlap: Each overlay tries to perform protection Inefficient: IP over ATM over SONET over WDM 22% Overlays do not often work in concert: every overlay runs at its own speed slower devices cannot fill the bandwidth No automated provisioning at the optical overlay the architecture does not scale well with increasing demand SONET Add and Drop Multiplexers: - Inflexible: Bit rate and signal format dependent - Costly: Introducing new OC-level requires new equipment
New Transport Network Architecture + Motivations Part 1 6 All-Optical DWDM Transport Networks Introduces concurrency to unlock The 50 Tbps potential capacity of single-mode fiber (i.e., bridging the opto-electronic bandwidth mismatch) Signal Format Independent (OC/STS/STM/Ethernet) Signal Bit Rate Independent (OC-Level Independent) Automatic Provisioning of light-paths
New Transport Network : Architecture (IP over DWDM) Part 1 7 Services Layer IP Routers Voice Switches DS-3 W-DCS Layer W-DCS W-DCS W-DCS OC-12 DS-3, OC-3, OC-12 B-DCS Layer B-DC S B-DC S B-DC S DS-3, OC-3, OC-12 Sonet ADM Layer ADM ADM ADM OC-48, OC-192, OC-768 Phtonic Layer (All-Optical DWDM transport Switches) OTS OTS OTS
New Transport Network Architecture : Advantages Part 1 8 Reduce the functional overlap between the overlays Independence from the signal formats and data rates Automatic provisioning of optical circuits More efficient transport of services (specially IP-based)
All-Optical DWDM Transport Networks Part 1 9 DWDM Types of Services: Permanent Optical Circuits (POC) Switched Optical Circuits (SOC) General Characteristics: Optical-Optical-Optical (O-O-O) Switches. No optical-electrical-optical (O-E-O) within the network. Dense Wavelenegth Division Multiplexing (DWDM Links).
All-Optical Transport Network : Protocols Part 1 10 ATM Switch UNI A NNI B NNI ATM Switch C UNI Signaling protocol: OIF-LDP or OIF-RSVP on UNI side GMPLS signaling for NNI Routing: GMPLS Routing
Routing and Wavelength Assignment : RWA Problem Part 1 11 Optical Paths that need to be established on O-O-O DWDM networks have to be assigned: Route: Path from source to destination thru intermediate nodes Wavelength(s): The lambda(s) assigned on each link thru the path Two variations of the RWA problem have been studied in the literature: Without wavelength continuity constraint (Wavelength Conversion expensive) With wavelength continuity constraint We focus on RWA problem in networks with Sparse wavelength conversion capabilities (because wavelength converters are expensive)
RWA Example : Wavelength Continuity Part 1 12
RWA Example : Wavelength Conversion Part 1 13
ILP Formulation : Approach Part 2 14 RWA problem is an NP-Complete problem The RWA problem has been formulated as an Integer Linear Programming (ILP) problem in the literature Our objective is to relax some of the constraints presented in previous literature to introduce an ILP that handles networks with Sparse and Limited Wavelength Conversion Capabilities We also introduce a pruning strategy that minimizes the number of variables and constraints in the ILP problem ILP can be applied to larger networks
ILP Formulation: RWA-SWC Part 2 15 The objective function of this formulation is to minimize the total cost of all lightpaths that need to be established in the optical network. Minimize Π i = ( Y subject to the following constraints: 1 i ) T S i S Q i j i 1 1 i Π 1 j ri r i S j= 1 i i j 1 i Π i T ( S ) U W 1 i Π i T i ( S ) X V 1 i Π i i T i ( S ) Z 1 1 i Π (1) (2) (3) (4) (5)
ILP Formulation: Pruning Strategy Part 2 16 In order to use our ILP formulation with larger networks We need to reduce the search space (non-optimal solutions) Pruning Strategy: (1) Limit the possible routes between source-destination pairs (2) Limit the possible wavelengths according to used tunable-lasers (3) Limit the possible wavelengths according to used wavelength converters (4) Limit the possible wavelengths to be the same before and after any node that does not support wavelength conversion (5) Limit the possible wavelengths to be a subset of the wavelengths that the DWDM links can support
ILP Formulation: Problem Size Part 2 17 Number of Variables = Number of Constraints = [(1 P ) W ] H * K *(1 P1 )* 2 * 4 + * *(1 P1 )* (1 P2 [ ) W ] H K * Number of lightpath requests K Average number of possible routes W Number of wavelengths per DWDM Link H Average number of hops P 1 Percentage of wavelength options pruned due to technology limitations P 2 Percentage of wavelength options pruned due to user s educated decision to use subset of the DWDM wavelengths
ILP Formulation: Problem Size (Cont d) Part 2 18 2500 2000 Count 1500 1000 500 0 0 5 10 15 20 25 30 Lightpaths Variables Constraints
ILP Formulation: Problem Size ( Cont d) Part 2 19 Count 45000 40000 35000 30000 25000 20000 15000 10000 5000 0 0 20 40 60 80 100 120 140 W Variables Constraints
ILP Formulation: Problem Size ( Cont d) Part 2 20 2500 2000 Count 1500 1000 500 0 0 0.2 0.4 0.6 0.8 1 P1 Variables Constraints
ILP Formulation: Problem Size ( Cont d) Part 2 21 2500 2000 Count 1500 1000 500 0 0 0.2 0.4 0.6 0.8 1 P2 Variables Constraints
ILP Formulation: Problem Size ( Cont d) c = 1 B 1 c = 0 D 1 1 A 1 1 F c = 2 c = 1 Scenario Source Destination Route Wavelengths Converters A B A F AB BD DF 3 3 3 0 0 A F AC CE 3 3 3 0 0 EF E A EC CA 1 1 0 E A EC CA 2 2 0 E B ED DB 2 2 0 C D CB BD 1 1 0 E A EC CA 3 3 0 E B ED DB 1 1 0 E B ED DB 2 2 0 E B ED DB 3 3 0 E B EC CB 1 1 0 E B EC CB 3 3 0 A B AB 1 0 A B AB 2 0 A B AB 3 0 A B AC CB 1 3 1 1 C c = 2 1 E c = 3 C : Number of wavelength converters installed on the node Number of wavelengths = 3 (per each bi-directional link) Part 2 22 1
ILP Formulation: Summary Part 2 23 Our ILP can handle networks with Sparse and Limited wavelength conversion capabilities Our ILP can be used to solve SLE problem (Typically : 6,000 variables and 6,000 constraints) Our Pruning strategy can be used to minimize the number of variables and constraints involved in the ILP problem handling bigger SLE problems
OSPF Extension : Introduction Part 3 24 K. Kompella, Y. Rekhter, "Routing Extensions in Support of Generalized MPLS," Internet Draft, Work in Progress, August 2002 K. Kompella, Y. Rekhter, "OSPF Extensions in Support of Generalized MPLS," Internet Draft, Work in Progress, August 2002. Our stand on GMPLS: Generic signaling and routing approach for PSC, TSC, LSC, FSC equipment. This introduces unnecessary complexities while ignoring the overlay model Did not address many of the details of all-optical DWDM transport networks (Advertising wavelengths, converters, origination policies, RWA techniques)
OSPF Extension : Our Approach Part 3 25 We need to know the availability of the wavelength and converter resources in order to perform SMART RWA or we will crankback or block the calls. Overlay-specific protocols are more efficient than generic ones. Overlay-specific protocols are less complex than generic ones. Having protocols that handle all the particularities of a specific switching technology efficiently is better than having a protocol that handles all switching technologies inefficiently. In this research we address the issues that GMPLS routing did not handle: (1) Advertising resource availabilities: Wavelengths and converters (2) Origination policies convenient for the dynamic nature of all-optical DWDM networks (3) RWA heuristics to conserve resource usage and minimize blocking
OSPF Extension : New Wavelength LSA Part 3 26 Link ID sub-tlv Local Interface sub-tlv Remote Interface sub-tlv Outgoing Interface sub-tlv Outgoing Interface sub-tlv Protection Type sub-tlv SRLG sub-tlv Wavelength Availability sub-tlv LS Age Option LS Type=10 TE Type TE LSA ID LSA# Advertising Router LS Sequence Number LS checksum Length=108 Type=2 Length=84 Type=2 Length=4 Link ID Type=3 Length=4 Local Interface IP Address Type=4 Length=4 Remote Interface IP Address Type=11 Length=4 Outgoing Interface Identifier Type=12 Length=4 Incoming Interface Identifier Type=32773 Length=4 Link Protocetion Type Not Used Type=32774 Length=8 Shared Risk Link Group (SRLG1) Shared Risk Link Group (SRLG2) Type=32775 Length=20 Length of Mask Wavelength Availability Mask LSA Header Link TLV Not Used Wavelength Resource availability Opaque LSA
OSPF Extension : New Conversion LSA Part 3 27 LS Age Option LS Type=10 TE Type TE LSA ID LSA# Advertising Router LS Sequence Number LS checksum Length=36 Type=32776 Length=12 LSA Header Converter Availability TLV Converter Type(1) Total Used Not Used Converter Type(2) Total Used Not Used Converter Type(3) Total Used Not Used Wavelength-Conversion Resource availability Opaque LSA
OSPF Extension : DWDM Simulation Part 3 28 Typical 16-node Motro-core/ Long-haul network
OSPF Extension : Protocol Simulation Part 3 29
OSPF Extension : Results Part 3 30 Blocking Probability 5.00E-01 4.50E-01 4.00E-01 3.50E-01 3.00E-01 2.50E-01 2.00E-01 1.50E-01 1.00E-01 5.00E-02 0.00E+00 0 2 4 6 8 10 12 14 Traffic Load (Erlang) No Advertisements Advertisements
OSPF Extension : Results (Cont d) Part 3 31 Blocking Probability 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Degree of Wavelength Conversion (%) No Advertisements Advertisements
Origination Policies: Immediate Strategy: Advertise changes in wavelength and conversion availabilities as soon as they happen. Advantages Simple, good for not very dynamic networks Disadvantages Generates a lot of advertisements (overhead) if if if if (One of the neighboring routers changes to/from the FULL state) then Originate a new Router LSA (LS-age field of the one of router s self-originated advertisement > LSRefreshTime) then Originate a new instance of the LSA that just expired (wavelength availability on one or more of the outgoing links changes) then Originate new wavelength availability opaque LSA (wavelength-conversion resource availability on the switch changes) then Originate new wavelength-conversion availability opaque LSA Part 4 32
Advantages of Using Fuzzy-Logic Part 5 33 Rule-Based Fuzzy Inference Systems (FIS) employ linguistic variables that can be easily understood and so modified or extended by others. Many criteria can be incorporated to make smarter routing decisions. Fuzzy Inference Systems (FIS) can easily incorporate the different heuristic algorithms employed in Operating Systems field (e.g., First-Fit, Last-Fit, Best-fit, etc.)
Origination Policies: Fuzzy-Based Part 4 34 Strategy: Even if we do not advertise resource availabilities when the wavelength and conversion resources are Not highly utilized, the blocking probability will not increase. Advantages Simple, extensible approach that can handle very dynamic networks with less overhead. Rule-base for Fuzzy-Based LSA Origination if if (Bandwidth Utilization is low) and (Converter Utilization is low) then Update Frequency is slow (Bandwidth Utilization is high) or (Converter Utilization is high) then Update Frequency is fast
Origination Policies: Fuzzy-Based (Example) Part 4 35
Origination Policies : Results Part 4 36 Number of exchanged Messages 9000000 8000000 7000000 6000000 5000000 4000000 3000000 2000000 1000000 0 0 2 4 6 8 10 12 14 Traffic Load (Erlangs) Immediate Flooding Fuzzy FLooding
Origination Policies : Results (Cont d) Part 4 37 Call Blocking Probability 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0 2 4 6 8 10 12 14 Traffic Load (Eralngs) Immediate Flooding Fuzzy Flooding
Origination Policies : Results (Cont d) Part 4 38 14000000 Number of Messages 12000000 10000000 8000000 6000000 4000000 2000000 0 0 20 40 60 80 100 Degree of Wavelength Conversion (%) Immediate flooding Fuzzy flooding
Origination Policies : Results (Cont d) Part 4 39 Call blocking probability 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 20 40 60 80 100 Degree of wavelength conversion (%) Immediate flooding Fuzzy flooding
Conclusions Proposed an ILP formulation and pruning-strategy for RWA problem in networks with sparse wavelength conversion capabilities. Proposed OSPF extension that enhances the blocking performance of networks with sparse wavelength conversion capabilities. Proposed fuzzy-based LSA origination policy that drastically reduces the number of messages exchanged over the control plane without hindering the blocking performance.
Future Extensions Evaluate the performance of the proposed OSPF extension and fuzzy origination policy for optical packet switching (OPS) and Optical Burst Switching (OBS) networks. Use Neuro-Fuzzy approach to dynamically learn and tune the parameters of the membership functions used in our fuzzy origination policy.