1 / 17

Multi-layer Traffic Engineering for IP-over-Optical networks

Multi-layer Traffic Engineering for IP-over-Optical networks. IMEC. Introduction (1). Reconfiguration based on IP traffic/QoS measurements. Information exchange (cost metric). Topology Optical metric. IP-over-Optical Networks. B. IP router IP link. A. D. E. C. IP network.

tanaya
Télécharger la présentation

Multi-layer Traffic Engineering for IP-over-Optical networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multi-layer Traffic Engineering for IP-over-Optical networks IMEC

  2. Introduction (1) Reconfiguration based on IP traffic/QoS measurements Information exchange (cost metric) • Topology • Optical metric IP-over-Optical Networks B IP router IP link A D E C IP network ASON/ION B optical switch optical fiber (WDM) lightpath A D E C Dept. of Information Technology - Ghent University

  3. Introduction (2) 70 60 50 40 Cost 30 20 10 0 0 20 40 60 80 100 Load(%) • General overview of MTE scheme used assign costs 3 4 route 2 5 full mesh signaling 1 · route traffic IP traffic or · set up necessarylightpaths 1’ failure ASON Dept. of Information Technology - Ghent University

  4. Introduction (3) moderate loads LC LMR LC:MC MC • MTE IP link load depending cost function 20 70 # links 60 cost 15 50 LLT 40 10 Cost # links 30 HLT 20 5 10 0 0 0 20 40 60 80 100 Load (%) Dept. of Information Technology - Ghent University

  5. Optical metrics (1) IP traffic flow grooming taking intoaccountthe opticallayer 9 l 6 l 2 1 3 2 2 1 1 1 1 1 Dept. of Information Technology - Ghent University

  6. Optical metrics – example (2) Static metric No metric • Many long lightpaths • No correspondence tophysical topology • Follows physical topology better • Shorter lightpaths • Topology seems more ‘open’ • Effect of cost metric in MTE strategy on logical topology Dept. of Information Technology - Ghent University

  7. Optical metrics in MTE (3) • Multi-layer Traffic Engineering strategy comes with an IP-layer cost function to handle: • IP/MPLS routing • Logical topology reconfiguration • Introduction of optical metric allows to take optical resources into account in setting up the logical topology • Optical metric used: C + S.#hops • calculated for each IP router pair (metric is used in IP/MPLS MTE) • #hops = optical hops for a lightpath between two IP routers. • Setting C and S allows to set the degree of optical resource optimization. • E.g. higher S = higher optical resource optimization  higher IP grooming Dept. of Information Technology - Ghent University

  8. Optical metrics in MTE (4) 70 60 50 40 Cost 30 20 10 0 0 20 40 60 80 100 Load (%) • Optical metrics in the MTE IP cost function • Multiply existing IP costfunction with optical metric • IP cost : grooming, topology • Optical : optimize optical layer • Note: optical cost as seen from • the IP layer, so each node pair • has an optical cost IP Optical ??? Dept. of Information Technology - Ghent University

  9. Optical metrics – costs (5) Optical metrics 12 C = 1 (flat) 2 C = 1, S = 0 (flat) C = 0.75 C = 0.75, S = 0.25 10 C = 0 C = 0, S = 1 C = –1, S = 2 C = –1 8 Same cost for single hop lightpath 1 optical cost optical cost 6 Cost penalty per lightpath 4 No fixed cost 0 0 0,5 1 1,5 2 Cost ‘bonus’ per lightpath 0 -1 optical hops 1 2 3 4 5 6 optical hops Cost = C + S.n Dept. of Information Technology - Ghent University

  10. Optical metrics – performance (6) maximum used wavelenghts on one fiber 18 flat (no metric) 16 C = 0.75 14 12 1 10 # lambda 8 6 2 • Better choice of lightpaths • Better (more) grooming 4 2 0 0% 10% 20% 30% 40% 50% Max IP load (<-> node pair) Maximum resource usage on single fiber C = 0 C = –1 Dept. of Information Technology - Ghent University

  11. Optical metrics – performance (7) slope 1 vs. no optical metric (relative) Relative optical savings (C = 0 vs. no metric) 60% total #Lambda max #Lambda 50% 40% 30% relative savings 20% 10% 0% 10% 15% 20% 25% 30% 35% 40% 45% 50% Max IP load (<-> node pair) Dept. of Information Technology - Ghent University

  12. Optical metrics – conclusion (8) optical metricsweet spot operatoradjustable Optical layer vs. IP layer Optical load vs. IP load 120 60 58 # Lambda 100 average IP load 56 80 54 load (%) #Lambda 60 52 50 40 48 20 46 0 44 1 0.75 0 –1 C (fixed penalty per lightpath) Dept. of Information Technology - Ghent University

  13. Optical metrics – conclusion (9) • Multi-layer Traffic Engineering strategy based on cost-function depending on IP link load. • Introduction of optical metric into the cost function to optimize the optical layer. • Even minimal information exchange (# hops) between optical and IP layer yields large improvement in optical layer utilization. • Simple optical cost metric with adjustable slope / fixed cost per lightpath allows a compromise between IP performance and optical layer cost. Dept. of Information Technology - Ghent University

  14. Prediction Based Routing (1) • Restraints such as wavelength continuity necessitate a proper routing strategy in the optical layer. • Not routing lightpaths, but typically also wavelength (and fiber) assignment: RWA. • Prediction Based Routing can select routes, wavelengths, fibers based on prediction (cf. branch prediction). • Find ways to combine the MTE strategy (mostly IP/MPLS layer) with the PBR strategy (optical layer). Dept. of Information Technology - Ghent University

  15. Prediction Based Routing (2) dest PBR decides on route, fiber, wavelength: • keeps history for each route-wavelength-fiber • selects from route x wavelength x fiber space using prediction fiber? route? wavelength? source Dept. of Information Technology - Ghent University

  16. Prediction Based Routing (3) • History for each Route/Fiber/Wavelength triplet (a ‘path’): • History becomes an index into a prediction table (separate table for each path). • Value in table[history-index] determines whether this path can be selected during lightpath set up • Important: construction of prediction table values is done completely at the source node, based on lambda utilization and blocking events: no information needs to be flooded to other nodes, lightpath setup uses only this local information! shift and update periodically … 0 1 0 past recent State means e.g. ‘path is utilized’ Dept. of Information Technology - Ghent University

  17. Prediction Based Routing (4) • The PBR strategy was developed by the Universitat Politècnica de Catalunya. In a joint effort with IMEC, the PBR was implemented using the MTE software tool. • Goals: • For MTE generated traffic loads, compare the performance of PBR against e.g. a first-fit strategy in the optical layer. note: MTE used fixed optical route/first free wavelength in the optical layer up till now. • Incorporate optical metrics that depend on optical load – we intend to extract an optical metric for each node pair from the PBR data structures. note: a lot of flooding in the optical layer lead to high overheads in optical load metric calculation. Dept. of Information Technology - Ghent University

More Related