R esearch and E xperimental A ssessment of C ontrol plane archi T ectures for I n- O peration flexgrid N etwork re-optimization (REACTION) Filippo Cugini, Luis Velasco, Juan Pedro Fernandez-Palacios Copenaghen, November, 2014
Outline • Project overview • Implemented project tools and facilities – OPNET model – Distributed testbed • Some project achievements on advanced use cases – Slice-ability – After failure repair optimization – Multipath restoration and bitrate squeezing 2
Participants Participant Participant Participant organisation name Country no. short name 1 Consorzio Nazionale Interuniversitario per CNIT Italy (Coordinator) le Telecomunicazioni Universitat Politècnica de Catalunya, 2 UPC Spain Barcelonatech 3 Telefonica TID Spain 3
Abstract • REACTION targets the design and validation of a flexible optical network enabling software-controlled super- channel transmission . • The focus is on: 1. Advanced bandwidth variable transponder (BVT) functionalities supporting multi-carrier transmission, Activity lead by CNIT adaptation of transmission parameters (mod. format, spectrum allocation, coding/FEC) and s lice-ability 2. Advanced control plane architecture and functionalities Activity lead by including innovative two-level active stateful PCE TID including the BGP-LS advertising solution 3. Advanced routing and spectrum assignment (RSA) Activity lead by UPC algorithms 4
Tool: OPNET Model C programming language OPNET: Event driven network framework • 5
OPNET Implementation (1/3): OSPF-TE Functionalities: • Link state advertisement • Spectrum availability adv. • Extensions for flex-grid • Routing and Spectrum Assignment • 6
OPNET Implementation (2/3): RSVP-TE Functionalities: • Reserve/Release spectrum slots along the path • Extensions for flex-grid • 7
OPNET Implementation (3/3): PCE and PCEP Finite state machine (FSM) of the FSM of the child process: specific for root process each PCE-PCC session. 8
Distributed control plane implementation • Extended for flexi-grid 9
Implemented control plane architecture Front-end PCE Back-end PCE Provisioning PCEP Active Solver PCEP PCEP Inventory Server BGP-LS Server TED LSP-DB TED LSP-DB • Active stateful front-end PCE , in charge of computing RSA and elastic provisioning • Back-end PCE responsible for performing complex network operation s (e.g., re-optimizations) • To provide the back-end PCE with updated network topology info, we propose to rely on the North-Bound Distribution of Link-State and TE Information through BGP , known as BGP-LS . 10
Testbed resources CNIT: - IP/MPLS network testbed composed of six PCE TED P 3 (QP QPSK) IP/MPLS routers (Juniper M7i/M10, Cisco 7206); - traffic generator/analyzer, OXC3 p 2 OXC4 - 3 ROADMs and 3 flexi-grid WSS; OXC5 - Optical terabit/s TX with coherent RX Tb/s Tb/s Tx/Rx Tx/Rx - Advanced control plane including GMPLS and OXC1 p 1 (16Q 16QAM) OXC2 PCE as well as SDN controller and agents. UPC: – Back-end PCE The test-bed supports the most advanced RSA algorithms and interoperates with the PCEs of the other partners. Telefonica I+D: - Flexgrid testbed with Flexgrid ROADMS from Cisco - control plane emulator supporting multiple domains and including GMPLS and PCE extensions for Flexgrid. 11
Slice-ability (1/2) Year 1 Year 2 Year 3 Four different destinations Two different destinations Single destination 37,5GHz Sliceable 200G 200G 400G 100G 100G 100G 100G functionality applied to a single four- carrier SBVT 150 GHZ 125 GHZ 100 GHZ • Flexibility to cope with traffic increase – support connections to different destinations , each served by a sub-set of sub-carriers. 12
Slice-ability (2/2) Year 1 Year 2 Year 3 Four different destinations Two different destinations Single destination 37,5GHz 200G 200G 400G 100G 100G 100G 100G 150 GHZ 125 GHZ 100 GHZ • Flexibility in provisioning and recovery d s Single path with low spectrum use vs. multi-path with overall larger spectrum? 13
Performance evaluation: Provisioning Sliced No Sliced Adaptive 14
Performance evaluation: Recovery Sliced Adaptive No Sliced [Ref] M. Dallaglio, A. Giorgetti, N. Sambo, F. Cugini, P. Castoldi, “Impact of slice-ability on dynamic restoration in GMPLS-based Flexible Optical Networks” Optical Fiber Communications Conference (OFC), March 2014 15
Slice-ability on different SBVT architectures • Sliceable BVT architecture can be implemented either with 1. Array of N tunable lasers 2. A single tunable multi-wavelength source (generated from 1 laser) • N lasers guarantee full and independent tunability of each sub-carrier no constraint s on RSA • Conversely, a MW source only supports contiguous frequencies RSA constraints • A MW source is expected to be to be cheaper (reducing the number of lasers), with lower footprint and lower power consumption. • Moreover a MW source enables better frequency stability among sub-carriers less spectrum. 16
Performance evaluation: Recovery with MW [Ref] M. Dallaglio, A. Giorgetti, N. Sambo, P. Castoldi, “Impact of SBVTs based on Multi-wavelength Source During Provisioning and Restoration in Elastic Optical Networks” , ECOC Conf, Sept. 2014 17
After failure repair optimization 4 slices 4 slices Re-optimization • 2 • 3 • 4 • 2 • 3 • 4 • 2 • 3 • 4 4 slices Restoration P1 P1 4 slices P1 1 • 6 • 7 5 1 • 6 • 7 5 1 • 6 • 7 5 8 slices 8 slices 2 slices • 1 • 1 P2b • 8 • 9 • 10 • 8 • 9 • 8 • 9 0 0 P2 P2 P2a • a) • b) • c) • Frequency slice • Frequency slice • Frequency slice 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1-2 1-2 1-2 2-3 2-3 2-3 P1 P1 P2b P4 P4 P4 3-4 3-4 3-4 4-5 4-5 4-5 • Optical Link 1-6 1-6 1-6 P2 P1 P2 6-7 6-7 6-7 7-5 7-5 7-5 1-8 P2b 1-8 1-8 P5 P5 P5 6-8 6-8 6-8 P2 P2 7-10 7-10 7-10 P2b 5-10 5-10 5-10 P5 P5 P5 8-9 8-9 8-9 P3 P3 P3 P2a 9-10 9-10 9-10 18
Effective re-optimization algorithms 19
Implementation (1/2) Telefonica CNIT UPC • ABNO-driven re-optimization • Involves both front-end PCE and Back-end PCE [Ref] L. Velasco, F. Paolucci, Ll. Gifre, A. Aguado, F. Cugini, P. Castoldi, V. Lopez,, “First experimental demonstration of ABNO-driven in-operation flexgrid network re-optimization” , OFC Conf, March. 2014, post-deadline paper 20
Implementation (2/2) Telefónica Premises UPC Premises (Madrid, Spain) (Barcelona, Spain) ABNO Back-end PCE Controller Active Solver PCEP 172.16.50.2 Server 172.16.104.2 PCEP TED LSP-DB PCEP 172.16.101.3 CNIT Premises Active Stateful PCE (Pisa, Italy) 2 Provisioning PCEP GCO Server TED LSP-DB PCC n = 1 m = 1 Res. Mngr. Conn. 10.0.0.49 Controller PCEP 10.0.0.8 10.0.0.1 Controller Controller [Ref]. L. Gifre, F. Paolucci, L. Velasco, A. Aguado, F. RSVP-TE Cugini, P. Castoldi, V. Lopez, Controller “ First Experimental Assessment of ABNO-driven In-Operation Flexgrid Network Re-Optimization ” Journal of Lightwave Technology (JLT), 2014. 21
Multipath Restoration and Bitrate Squeezing 22
SDN-based Implementation [Ref]. F. Paolucci, A. Castro, F. Cugini, L. Velasco, P. Castoldi “ Multipath restoration and bitrate squeezing in SDN-based elastic optical networks [Invited] ” Journal of Photonic Network Communications (PNET), Aug. 2014 23
Multi-domain networks Broker Multi-domain planning tool SDN Controller SDN Controller SDN Controller ABNO In-operation In-operation Planning Tool Planning Tool Inter-domain Links Abstract Links DataCenter 1 DataCenter 2 Optical Transport Optical Transport Network Network 24
Example Candidate Path src Domain 2 (ABNO) 1, 3, 6 1, , 6 Domain 1 1, 2,6 1, ,6 1, 5, 6 1, , 6 free s e slices es: 1, 1, 3, 3, 4, 6 , 6 3, 6 , 6 tgt 1, 6 , 6 1, 5 , 5, 6 , 6 Domain 4 25
Experimental Set-up • 1 • 6 USTC Premises Broker (Hefei, China) 169.237.74.168 OF-Controller UPC Premises D2 (Barcelona, Spain) HTTP/XML OpenFlow OpenFlow 147.83.30.189 222.195.92.10 PLATON OpenFlow HTTP REST Server Controller D1 Controller UC Davis Premises 169.237.74.223 (Davis, California) OpenFlow Topology • 9 DB Algorithm • 8 • 1 • 2 • 3 • 4 • 5 • 6 • 7 • 8 • 9 • 10 [Ref] Ll. Gifre et al, " Experimental Assessment of Broker and Planning Tool Coordination in Multi-domain Environments ,“ submitted to OFC, 2015. 26
Recommend
More recommend
