1. NOBEL WP5 Meeting Munich – 14 June 2005 WP5 Cost Study Group Author:Martin Wade (BT) Lead:Andrew Lord (BT) Relative Cost Analysis of Transparent & Opaque Optical Network Design Solutions

2. 2 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Objectives  Is the transparent optical network cheaper than the “traditional” network, and if yes: what are the boundary conditions?  Analyse the following scenarios: –Fully transparent networks (O/O/O, optical switch matrix, no transponders for transit traffic and thus no wavelength conversion) –Traditional opaque networks (O/E/O, electrical switch matrix, point-to- point links only, transponders required for transit traffic) –Hybrid networks (transparency islands with transponders for signal regeneration and/or wavelength conversion) –New vendor solutions, e.g. Infinera cheap integrated 10 * 10 Gbit/s O/E/O chips  Understand underlying design and cost factors –Consider ‘pragmatic’ & ‘realistic’ rather than optimised designs

3. 3 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Inputs  Reference networks –Pan-European, DT, TILAB, BT core & London metro  Traffic demand –Three traffic growth levels per network –10 Gbit/s granularity only –Traffic assumed to be already groomed Potential advantage of opaque solution thus not taken into account Client layer architectures not considered, although heavy impact on traffic profile –Unprotected Protection and shared restoration not yet considered –Static (once provisioned, it remains for long time) Dynamic traffic not yet considered  Network element costs –Capex only considered initially

4. 4 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Methodology  Excel/VBA calculations performed –No formal modelling tools available for general use  Routing for opaque and transparent designs –Fixed routing for each individual traffic demand Adaptive routing not yet applied –Least cost routing (minimum hop and minimum path distance other options)  Wavelength assignment for transparent design –‘First fit’ method applied (lowest numbered spare wavelength used; lowest numbered OMS used for each link) –Order of traffic assignment options: max hop paths first or random order (simulates incremental provisioning with no foresight)

5. 5 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Node Models Opaque Switch Line Cards & Transponders Opaque Switch Coloured Line Cards Transparent Switch

6. 6 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Network Cost Elements  Input from WP2 D15 with additional data from Siemens –Data to be verified  Major items used: –Base unit: Electrical STM-1 line-card 1 –STM-64 LR transponder160(too high) –STM-64 SR line card 70(only for add/drop) –40 channel mux/demux module220 –80 channel mux/demux module500 –Amplifier SH220 –Signal conditioning element430(transparent only) –Opaque switch per port 18.75(derived) –Transparent switch per port 12.5(derived)

7. 7 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade High-Level Design Comparisons  Switch port dimensioning –Transparent & Opaque2 * Traffic * (Av hops + 1)  Transponder dimensioning –Transparent2 * Traffic Additional for regeneration over transparency limit –Opaque2 * Traffic * Av hops Proportion of transponders for drop traffic1 / Av hops Proportion of transponders for transit traffic1 – (1 / Av hops) May be implemented as: –line card & transponder –coloured line card (probably cheaper, but there may be compatibility issues with using different OMS supplier)

8. 8 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade High-Level Design Comparisons  Line card dimensioning (Opaque only) –Tributary line cards2 * Traffic –Trunk line cards2 * Traffic * Av hops Only if line cards are used with transponders (see previous slide)  OMS channel dimensioning (per link) –OpaqueNo of connections on link –TransparentMax channel used on link –Ratio of (max channel / link load) is the Wavelength Continuity Penalty Results in more OMS channels required to carry all traffic or fewer routable traffic demands Dependent on network topology, OMS channel capacity, order of routing assignment –Sum of all opaque link loadsTraffic * Av Hops

9. 9 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Design Results – Pan-European  16 nodes; 23 links  Key stats  Total Costs (excl. ILAs) * Based on 40 channel OMS systems and traffic provisioned in random order (average of 100 runs)

10. 10 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Design Results - DT  17 nodes; 26 links  Key stats  Total Costs (excl. ILAs) * Based on 40 channel OMS systems and traffic provisioned in random order (average of 100 runs)

11. 11 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Design Results - TILAB  38 nodes; 79 links  Key stats  Total Costs (excl. ILAs) * Based on 40 channel OMS systems and traffic provisioned in random order (average of 100 runs)

12. 12 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Design Results – BT Core  79 nodes; 139 links  Key stats  Total Costs (excl. ILAs) * Based on 40 channel OMS systems and traffic provisioned in random order (average of 100 runs)

13. 13 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Design Results – BT Metro  38 nodes; 62 links  Key stats  Total Costs (excl. ILAs) * Based on 40 channel OMS systems and traffic provisioned in random order (average of 100 runs)

14. 14 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Initial Conclusions  As suspected, costs of transponders in opaque designs far outweigh additional OMS costs in transparent designs –Opaque designs 1.9 – 2.5 times more expensive than equivalent transparent designs  However, transponder costs are clearly far too high in current cost model –Effect of Infinera cheap integrated 10 * 10 Gbit/s O/E/O chips also to be determined, e.g. what fraction would they have to be to cost in?  Effect of loading traffic in random order compared to near-optimal heuristic ranges from 2% to 40% additional OMS channels required  Smaller OMS system capacities significantly reduce the additional OMS channels required - up to 88% more if there is no OMS capacity limit - due to effective wavelength conversion with multiple stacked OMS systems on same link

15. 15 NOBEL WP5 Ipswich Meeting - 18 March 2005BT - Martin Wade Further Work  Cost sensitivity analysis  Adaptive (state-dependent) routing  Traffic protection and shared restoration  Trade-off of transparency limit vs. regenerator costs –Analysed by Matthias for DT network –Transparency limit (1500 km) currently so high very few traffic paths exceed limit –Will allow transparency islands to be analysed  Opex considerations –Reliability of components –Operation of network