OPTCOM GROUP DIPARTIMENTO DI ELETTRONICA E TELECOMUNICAZIONI POLITECNICO

OPTCOM GROUP DIPARTIMENTO DI ELETTRONICA E TELECOMUNICAZIONI POLITECNICO DI TORINO ITALY

MOTIVATION Btot Bstat Allocated bandwidth Supposed to be defragmented Bdyn Bandwidth available for dynamic spectral allocation Total available bandwidth • Most networks are running out of spectrum • New – and supposedly dynamic – traffic has to be allocated on the residual bandwidth • What is the advantage of flex- vs. fix-grid given the residual bandwidth? ICTON 2017, paper Th. B 4. 5 curri@polito. it 2

OUTLINE • The generalized OSNR as l Qo. T and the LOGO control plane and the statistical network assessment process (SNAP) • Residual bandwidth percentage • Traffic model • Fix- and fle-grid transceivers • The Italian network topology • Results as BP vs. Total allocated traffic • Comments and conclusions ICTON 2017, paper Th. B 4. 5 curri@polito. it 3

AN ALL OPTICAL TRANSPARENT NETWORK With the knoweledge of the network physical details D Example: Qo. T for lightpath travelling from node A to node G through nodes C and E: F E B A We set the optimal power per channel and manage the network as an inverse-OSNR (IOSNR) weighted graph G C H I ICTON 2017, paper Th. B 4. 5 where curri@polito. it 4

STATISTICAL NETWORK ASSESSMENT PROCESS i=i+1 Metrics’ evolution vs. j j=j+1 Traffic model i-th Monte Carlo run Start J-th Traffic request RWA algori thm Network description NO sat ura tion RWA NO YES I> NMC YES SNAP Network snap-shot @ i, j Network evolution storage Static Metrics M. Cantono et al. “Potentialities and Criticalities of Flexible-Rate Transponders in DWDM networks: a Statistical Approach , ” JOCN, 2016 V. Curri et. al. , “Elastic All-Optical Networks: a New Paradigm Enabled by the Physical Layer. How to Optimize Network Performances? , ” JLT, 2017 ICTON 2017, paper Th. B 4. 5 curri@polito. it 5

RESIDUAL BANDWIDTH PERCENTAGE (RBP) Bstat Allocated bandwidth Supposed to be defragmented Bdyn Bandwidth available for dynamic spectral allocation Btot: C-band • Total available bandwidth: C-band of 4000 GHz • RBP=Bdyn/Btot x 100 [%] • We consider three different scenarios 1. Lightly loaded network: RPB = 25% 2. Moderately loaded network: RBP = 50% 3. Heavily loaded network: RBP = 75% ICTON 2017, paper Th. B 4. 5 curri@polito. it 6

TRAFFIC MODEL • Any-to-any traffic • Flat probability in traffic requests • Traffic requests «groomed» at • RG = 20 Gbps • RG = 40 Gbps • RG = 100 Gbps • 5000 Monte-Carlo progressive network loading up to blocking probability BP of 30% • RSWA algorithm: shortest path up to kmax=50 ICTON 2017, paper Th. B 4. 5 curri@polito. it 7

FIX AND FLEX GRID TRANSCEIVERS Fix-grid Flex-grid f. LPi+1 f. LPi+2 f. LPi+3 f. LPi+4 f • WDM grid: Df =37. 5 GHz f • Grid: Df=RS, slot =12. 5 Gbaud, up to 5 slots • RS =31. 25 Gbaud • Coding overhead = 25% • fch tunable on 6. 25 GHz • Guard-band: = 6. 25 GHz • Guard-band = 6. 25 GHz Modulation Format Net Bit rate (R b) Gbps Modulation Format PM-BPSK 50 PM-QPSK Net Bit rate (R b) Gbps Ns=1 Ns=2 Ns=3 Ns=4 Ns=5 PM-BPSK 20 40 60 80 100 PM-QPSK 40 80 120 140 280 PM-16 QAM 200 PM-16 QAM 80 160 240 320 400 PM-64 QAM 300 PM-64 QAM 120 240 360 480 600 ICTON 2017, paper Th. B 4. 5 curri@polito. it 8

THE ITALIAN NETWORK TOPOLOGY ICTON 2017, paper Th. B 4. 5 curri@polito. it 9

RESULTS • Metrics • Normalized Network Traffic (T) is the total traffic loading the network normalized with respect the available bandwidth: it is indeed a specrat eexploitation efficiency • Blocking probability (BP): it is the probability that a new traffic request is rejected • Results are shown as BP vs. T for the different scenarios • Comparisons at BP = 1% are shown as well ICTON 2017, paper Th. B 4. 5 curri@polito. it 10

BP VS. T: FIX-GRID Blocking Probability [%] 30% 10% __ R G= 20 G __ R G= 40 G __ R G=100 G 1% 0 … RBP= 25% _. _RBP= 50% _ _RBP= 75% 5 10 15 20 25 Normalized Network Traffic [Tbps/THz] ICTON 2017, paper Th. B 4. 5 30 35 curri@polito. it 11

BP VS. T: FLEX-GRID Blocking Probability [%] 30% 10% __ R G= 20 G __ R G= 40 G __ R G=100 G 1% … RBP= 25% _. _RBP= 50% _ _RBP= 75% 0. 1% 0 5 10 15 20 25 Normalized Network Traffic [Tbps/THz] ICTON 2017, paper Th. B 4. 5 30 35 curri@polito. it 12

25 20 FIXED GRID FLEX GRID 15 11. 6 8. 8 10 5 16. 5 3. 6 4. 4 5. 0 0 25% 50% 75% Residual Bandwidth [%] RG = 40 Gbps 30 25 20 FIXED GRID 21. 1 24. 2 17. 3 15 10 7. 1 8. 7 9. 7 5 0 25% 50% 75% Residual Bandwidth [%] Normalized Network Traffic [Tbps/THz] RG = 20 Gbps 30 Normalized Network Traffic [Tbps/THz] FLEX VS. FIX @ BP=1% 30 25 20 RG = 100 Gbps FIXED GRID FLEX GRID 21. 2 24. 3 19. 2 15 10 5 0 25% 50% 75% Residual Bandwidth [%] • RBP=25% • RBP=50% • RBP=75% • Flex 2 Fix 2. 5 • Flex 2 Fix 3. 3 • Flex 2 Fix 2. 5 ICTON 2017, paper Th. B 4. 5 20. 2 16. 7 • RBP=25% • Flex 2 Fix 2. 5 26. 1 • Flex 2 Fix 1. 3 curri@polito. it 13

CONCLUSIONS • We showed that flex-grid is always advantageous with respect to fix-grid • The relative advantage that is around 2. 5 times for grooming sizes of 20 and 40 Gbps, decreases to 1. 3 times for RG= 100 Gbps because of saturation effects • Spread of spectral efficiency for different scenarios of RBP and RG is much smaller for flex-grid, while using fix-grid the larger RG, the better ICTON 2017, paper Th. B 4. 5 curri@polito. it 14

This work was supported by CISCO Systems within a SRA contract ICTON 2017, paper Th. B 4. 5 curri@polito. it 15