Routing and Wavelength Assignment in Optical Networks

1. Routing and Wavelength Assignment in Optical Networks Xavi Masip, Sergi Sánchez, Eva Marín and Josep Solé, UPC {xmasip, sergio, eva, pareta}@ac.upc.edu

2. Outline • Introduction • Routing in OTN • Signalling in OTN • Wavelength conversion • QoS in OTN • New routing proposals • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

3. Introduction • Control Plane • Routing issues: compute the lightpath • Signalling issues: establish the lightpath • GMPLS1 is a strong candidate • Routing protocols • Constraint-based routing • Signalling protocols • RSVP-TE2 • CR-LDP3 1L.Berger, “Generalized Multi-Protocol Label Switching (GMPLS): Signaling Functional Description”, IETF, RFC3471, Jan., 2003. 2 “GMPLS signalling resources reservation protocol-traffic engineering (RSVP-TE) extensions”, IETF, RFC3473, Jan. 2003. 3 P.Ashwood-Smith et al, “GMPLS signalling constraint-based routed label distribution protocol (CR-LDP) extensions”, IETF, RFC3472, Jan.2003.

4. Introduction QoS Routing • Routing in IP networks • QoS constraints • Source routing (CBR) BW, delay, jitter, loses,… Network state information ?

5. Introduction • Routing target • Establish a lightpath between source-destination node pair • Routing open issues • Depends on network: • Circuit-switched (ASON) • Burst-switched (OBS) • Packet-switched (OPS) • Depends on wavelength conversion capabilities • Depends on control plane • Centralized lightpath establishment • Distributed lightpath establishment (source/destination-based) • QoS routing: what does it mean in OTN? • Network parameters • Update policies • …

6. Introduction Cables Fibers Wavelegnths Bands Wavelengths {l..l}1 l TimeSlots Packets t0 t1 t2

7. Outline • Introduction • Routing in OTN • Routing constraints • The RWA problem • Signalling in OTN • Wavelength conversion • QoS in OTN • New routing proposals • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

8. Routing in OTN ASON, OBS, OPS • Routing and wavelength assignment problem • Determine both: physical route and wavelength/s to be assigned • Connection management protocol Physical path, wavelength assignment

9. Routing Constraints • Routing depends on connection management • In legacy backbone networks traffic is generally static • In next-generation optical networks traffic is more dynamic • OBS • Connection requests arriving at high rates • Average duration about several tens or hundreds of milliseconds • New dynamic lightpath provisioning schemes • Control plane is required: GMPLS

10. Routing Constraints • Wavelength continuity-constraint • Wavelength converters are expensive (at least for the next 5 years) • Wavelength Selective networks (WS): • There are not wavelength conversion devices • Same wavelength on all the route • Wavelength Interchangeable networks (WI): • There are wavelength conversion devices • Different wavelengths in the end-to-end path

11. Routing Constraints • Traffic in a wavelength-routed network • Static traffic pattern • Lightpaths set up all at once • Remain in the network for a long period of time • Dynamic traffic pattern • Lightpath is set up reacting to an incoming request • Lightpaths take down dynamically • Dynamic lightpath (on-demand) is foreseen to best accommodate current and future Internet traffic • Big challenge • Develop efficient algorithms and protocols for establishing lightpaths

12. The RWA problem • Dynamic lightpath establishment • Routing and wavelength assignment problem (RWA) • It is NP-complete to find the optimal solution for the two problems1 • Addressed separately • Through heuristic algorithms • Routing decomposed into: • Routing problem • Wavelength assignment problem 1A.Mokhtar, M.Azizoglu, “Adaptive Wavelength Routing in all-optical networks”, IEEE/ACM Trans.Netw., 1998

13. The RWA problem • Routing subproblem • Static approaches • Fixed routing • A single fixed route is precomputed • Ex: Shortest Path • Fixed-alternate routing • Multiple fixed routes are precomputed • Ex: Fixed shortest-path routing1, Least-Loaded Routing 1 • Adaptive (dynamic) approaches • Ex: Link state approach2, distributed routing approach3 • Adaptive routing based on global information • Adaptive routing based on neighbourhood information • Adaptive routing based on local information 1 E.Karasan, E.Ayanoglu, “Effects of wavelength routing and selection algorithms on wavelength conversion gain in WDM optical networks”, IEEE/ACM Trans.Netw., 1998 2 B.Muhherjee, “WDM Optical Communication Networks: Progress and Challenges”, IEEE JSAC, 2000 3 H.Zang, et al, “Connection management for wavelength-routed WDM networks”, Globecom 1999

14. The RWA problem • Adaptive routing based on global information • Routing decisions based on network state information • Full information about wavelength availability • Implemented in: • Centralized manner • Distributed manner: • Standardized by GMPLS • Different implementations • Link state approach • Distance-vector approach (or distributed routing approach) • Least-congested path algorithm

15. The RWA problem • Centralized manner • A single device maintains complete network state information • A single device finds routes and establishes the lightpaths • No huge coordination between nodes is required • Critical point of failure • Distributed manner • Decisions are distributed to different network nodes • GMPLS-based network: • Routes are computed by any routing (RWA) algorithm • A signalling (connection management) scheme is responsible for establishing the lightpath • Signalling candidates: • RSVP-TE • CR-LDP • There is not any guarantee that the updated global information with respect to wavelength availability on each link will be available in a distributed environment

16. The RWA problem • Distributed manner • Link state approach: • Each node maintains complete network state information • Each node finds the route in a distributed manner • Network state updating whenever network state changes • Signalling overhead vs outdated information • Distance-vector approach (or distributed routing approach): • Each node does not maintain global network state information • Each node maintains a routing table indicating the next hop and the distance to the destination • Updating is also required • Least-congested path algorithm: • Congestion on a link = number of available wavelengths • Links with fewer available wavelengths = links more congested • Congestion on a path = congestion on the most congested link on the path • Selects a set of sequence of routes for each source-destination pair • The least congested is selected • Using SP + LCP performs better than only using LCP

17. The RWA problem • Summary • Routing schemes based on global network state information perform better whenever update network state information • Suitable for networks where lightpaths do not change much with time • Small optical networks • Static traffic (not bursty) • Good updating procedure (signalling protocol) • New routing schemes assuming outdated network state information

18. The RWA problem • Adaptive routing based on neighbourhood information • In the LCP all links on all paths must be examined in choosing the least congested one • Network state information must be either: • Maintained by each node • Gathered in real time when establishing the lightpath • LCP variant: • Only examines the first k-links on each path (neighbourhood information) • K = 2 behaves similar than fixed-alternate routing

19. The RWA problem • Adaptive routing based on local information (deflection routing) • Also named alternate link routing • Another approach to adaptive routing with limited information • Selects from alternate links on a hop-by-hop basis instead of alternate routes on a end-to-end basis • Each node maintains a routing table indicating one or more outgoing links to each destination • The outgoing links are precomputed and can be properly ordered • A link is selected depending on its wavelength availability • Each node only maintains information regarding the wavelength availability on its outgoing links • No update messages are required

20. The RWA problem • Dynamic routing seems better than static routing: • Considers network state information when selecting lightpaths • BUT • Traditional dynamic routing algorithms selects the path maximizing the number of available wavelengths • the length of the routes and the wavelength distribution must be jointly considered in the lightpath selection process • Dynamic routing + global information = routing inaccuracy • Shortest path selects routes only based on hop length • Fixed-alternate routing is a trade-off mechanism

21. The RWA problem • Wavelength Assignment subproblem • Random, First-Fit, Least-Used, Most-Used, Min-Product, Least-Loaded, Max-Sum, Relative Capacity Loss1 • The literature results show that the First-Fit2: • Very good performance • Is very simple to implement • Examples: • Fixed-alternate routing and first-fit wavelength assignment (FAR-FF) • Least-Loaded routing and first-fit wavelength assignment (LLR-FF) • Fixed-alternate vs dynamic • LLR-FF performs better • LLR-FF longer setup delays and higher control overheads 1H.Zang, et al., “A Review of Routing and Wavelength Assignment Approaches for Wavelength-Routed Optical WDM Networks”, ONM, 2000 2 Y.Zhu, et al, “A Comparison of Allocation Policies in Wavelength Routing Networks”, Photon. Net. Commun.J., 20000

22. Outline • Introduction • Routing in OTN • Signalling in OTN • Signalling schemes • Updating procedure • Wavelength conversion • QoS in OTN • New routing proposals • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

23. Signalling schemes • A signalling protocol is required to reserve resources along the selected route • Three signalling schemes • Source initiated reservation (SIR) • Wavelength resources are reserved as the control messages traverses along the forward path to the destination • Destination initiated reservation (DIR) • Connection request collecting wavelength availability on each link • Based on this information the destination node will select an available wavelength along the path • Intermediate-node initiated reservation (IIR) • Some intermediate nodes can initiate the wavelength reservation

24. Signalling schemes • SIR • Source routing (IP networks) • Wavelength/s are reserved as setup message is forwarded to the end • Wavelength number = f (wavelength information accuracy) • If source node has complete network state information: • May reserve only one wavelength • Complete network state information is not usual • Reserve all possible or a group of wavelengths • SIR problem: OVER-RESERVATION • Consequence: blocking of simultaneous connections

25. Signalling schemes • SIR

26. Signalling schemes • DIR • Blocking produced by outdated information • DIR outperforms SIR (no wavelength conversion)

27. Signalling schmemes • SIR problem • Over-reservation (many wavelengths are reserved) • DIR problem • Outdated information (only one wavelength is reserved) • IIR: proposed solution • Allows reservations to be initiated by a set of intermediate nodes before connection request arrives at the destination node • Reduces the over-reservation • Reduces the vulnerable period = outdated information

28. Signalling schemes • IIR without last link conflict

29. Signalling schemes • IIR with last link conflict

30. Updating procedure • Update in OTN • Updating is needed to refresh network changes • Information about topology and resource availability • Done by LSAs (link state advertisements) based on flooding • Dynamic network = huge amount of update messages = congestion • Triggering policies: • Update reduction • Inaccurate network state information • Triggering policies • Periodical updating • Relative change based triggers • Current link state – past link state > th (%) • Absolute change based triggers • Number of changes > th • Lazy flooding (not published yet) • Threshold, exponential, fibonacci

31. Outline • Introduction • Routing in OTN • Signalling in OTN • Wavelength conversion • Wavelength conversion schemes • Wavelength conversion trade-off • Wavelength converters placement • Signalling schemes and wavelength conversion • QoS in OTN • New routing proposals • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

32. Wavelength Conversion scheme • A lightpath connection could be blocked by • A route with a free available wavelength from s-d cannot be found • A wavelength cannot be found between s-d although there is free capacity on every hop of the path • Wavelength continuity constraint (WCC) • Blocking due to the WCC is dominant in the global blocking • Solution: Remove the WCC • Introducing wavelength converters • Problem: Wavelengths converters are expensive • It is not feasible to include a wavelength converters on all the network nodes • Not affordable in the next 5 years (CISCO said)

33. Wavelength conversion scheme l1 Fiber #1 Fiber #1 l100 lp l1 Fiber #N Fiber #N l100 The labels indicate wavelengths (or frequencies)

34. Wavelength conversion scheme Fiber #1 Fiber #1 Fiber #1 lp {l..l}10 Fiber #N Fiber #N {l..l}10 The labels indicate wavelength bands

35. Wavelength conversion scheme • Solution • Sparse wavelength conversion • Limited wavelength conversion • Sparse wavelength conversion • Wavelength conversion is available only on a set of nodes • New converters placement schemes must be developed • Limited wavelength conversion • Limits the range of conversion by a fixed value k • Translation degree D • Λ is the total number of wavelengths on a link

36. Wavelength conversion scheme • In short • In WDM networks with centralized wavelength provisioning, sparse conversion could achieve nearly the same performance as full wavelength conversion • The question is: • Given a network topology, a certain number of wavelength converters, and traffic statistics • How can the wavelength converters be placed into the network in order to minimize the overall blocking probability?

37. Wavelength converters placement • Easy way • Analyze nodes more congested • These nodes are candidates to get conversion capabilities • Best way: • There are many converter placement algorithms for simple topologies (bus, ring) • The complex the topology the complex the algorithm • Many heuristics are currently defined for static routing + random • It has been demonstrated that any converter placement algorithm does not work on any RWA1 • Both issues must be tackled together 1X.Chu, B.Li, I.Chlamtac, “Wavelength Converter Placement under Different RWA Algorithms in Wavelength-Routed All-Optical Networks”, IEEE ToN, vol51, nº4, April 2003

38. Wavelength converters placement • Target: • Place the wavelength converters in those nodes so that blocking decreases • Proposals1 • Minimum blocking probability first (MBPF) • To be applied to the FAR-FF (fixed-alternate routing) • Weighted Maximum Segment Length (WMSL) • To be applied to the LLC-FF (dynamic routing) • Boths solutions assume: • Sparse wavelength conversion • Full range conversion (not limited) • Based on placing converters one by one sequentially 1X.Chu, B.Li, I.Chlamtac, “Wavelength Converter Placement under Different RWA Algorithms in Wavelength-Routed All-Optical Networks”, IEEE ToN, vol51, nº4, April 2003

39. Wavelength converters placement • MBPF: • Find the N feasible paths according to the FAR • Define candidate nodes as those nodes without conversion capabilities • Assume a WC on all candidate nodes (one by one) • Compute blocking based on a mathematical model • Place a WC in that node minimizing the blocking • Repeat the process till the M WC are placed • WMSL: • Assign a weight to each candidate node indicating the impact of this node on the blocking • Assume that the hop length significantly contributes to the blocking for WS networks • Introduce WC so that the end-to-end paths are divided into several “WS” segments. • Blocking is fully dependent on the segment length

40. Signalling schemes and Wavelength conversion • DIR for networks with sparse wavelength conversion

41. Signalling schemes and Wavelength conversion • IIR for networks with sparse wavelength conversion

42. Outline • Introduction • Routing in OTN • Signalling in OTN • Wavelength conversion • QoS in OTN • New routing proposals • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

43. QoS in OTN • What does QoS mean in OTN? • The answer is not easy at all • What we know is how network performance can be evaluated • The blocking probability • Look at: • Customer needs • Applications requirements • Operator technical requirements • Service classes

44. Usability Requirements blocking probability network availability application set-up time connection set-up time network robustness throughput delay jitter packet loss rate BER (bit error rate) bandwidth requirements security requirements nomadism connectivity (number of senders/receivers) naming and numbering Operator Requirements (fulfill user requirements with the lowest possible quantity of resources and in the most efficent way) business roles elasticity (level of modification of original traffic shape) interactivity availability nature of bidirectional traffic (symmetric or asymmetric) provisioning of network resources naming and numbering supervision of network performance and faults (monitoring) AAA (Authentication, Authorization, Accounting) QoS in OTN

45. QoS classes emergency („strict“ real-time) conversational (real-time) streaming („almost“ real-time) prioritized elastic (business non-real-time) best effort (non-real-time) Optical network services connection-oriented analogue connection-oriented with wavelength switching connection-oriented burst switching connectionless burst switching connectionless packet switching transparent or opaque lightpaths dedicated or shared lightpaths Selection of QoS parameters required optical network service max. and min. holding times connection set-up and tear-down time blocking probability bandwidth requirements latency BER routing stability controlling and signalling delay load overhead for admin, mgmt etc. service availability in case of OBS service: burst aggregation mechanism, burst loss and blocking probability physical parameters in case of analogue optical network service QoS in OTN

46. Outline • Introduction • Routing in OTN • Signalling in OTN • Wavelength conversion • QoS in OTN • New routing proposals • MICORA (Minimum Coincidence Routing Algorithm) • PBR • The Routing Inaccuracy Problem • The Routing Inaccuracy Problem in hierarchical networks • Open issues

47. MICORA • Dynamic routing seems better than static routing: BUT • Traditional dynamic routing algorithms selects the path maximizing the number of available wavelengths • the length of the routes and the wavelength distribution must be jointly considered in the lightpath selection process • Dynamic routing + global information = routing inaccuracy • Shortest path selects routes only based on hop length • Fixed-alternate routing is a trade-off mechanism • New proposal • Trade-off between network performance and complexity • Fixed-alternate routing + wavelength assignment • MICORA (Minimum Coincidence Routing Algorithm) • Aims to optimize the physical route selection • Aims to reduce effects of selecting paths regardless network state • Does not select the k-shortest paths • Selects paths with minimum “coincidences”

48. MICORA • MICORA1 • Selects the k-(shortest & minimum coincident) paths • Computes the end to end paths considering the routes that have less shared links and minimum number of hops • Looks for the k-routes in two iterative basic steps • MICORA steps to compute k-routes • Selects the shortest path • Computes the Minimum Shared Link (MSL) MSL =NH * SL • NH: number of hops • SL: number of links shared between each path and the previous selected path • Select the path with minimum MSL • Repeat the last two steps (k-2) times 1S.Sánchez-López, X.Masip-Bruin, J.Solé-Pareta, “The Minimum Coincidence Routing in Optical Networks”, Workshop on G/MPLS, Girona, March 2006

49. Incoming request from node 1 to node 13 K = 3 MICORA steps: Selects the shortest path as the first path MICORA: Illustrative Example Shortest paths Selected paths 1-7-12-13 1-2-8-12-13 1-7-12-13 1-6-8-12-13 1-2-8-5-9-11-14-12-13 1-6-8-5-9-11-14-12-13

50. MICORA: Illustrative Example Path = 1-7-12-13 SL MSL Selected paths 1-2-8-12-13 1 4 * 1 = 4 1-7-12-13 1-6-8-12-13 1 4 * 1 = 4 1-2-8-12-13 1-2-8-5-9-11-14-12-13 1 8 * 1 = 8 1-6-8-5-9-11-14-12-13 1 8 * 1 = 8 • MICORA steps: • Computes the MSL parameter • Selects the path minimizing the MSL value • First iteration to compute the second path • Second iteration to compute the third path