Joint ITU-T/IEEE Workshop on The Future of Ethernet Transport Economics of the Next Data Rate Chris Cole 28 May 2010 Geneva Outline ■ Next Ethernet Data Rate Requirements ■ 100GE Transceivers ■ 400GE Transceivers ■ 1TE Transceivers ■ Summary 28 May 2010 2 Technology Product Life Cycle ■ ■ ■ Advanced (ex. 100GE in near term) ■ Major R&D investment ■ Moderate volume high end market ■ 10x rate Δ @ >30x cost Δ Parity (ex: 40GE in near term) ■ Moderate R&D investment ■ Medium volume multiple markets ■ 10x rate Δ @ 10x cost Δ Mature (ex: 10GE in near term) ■ Incremental R&D investment ■ High volume mass market ■ 10x rate Δ @ <4x cost Δ 28 May 2010 3 Next Data Rate Requirements ■ ■ Requirements from multiple developers ■ Leverage major R&D investment ■ Leverage ramping product volumes Requirements from multiple end users ■ Scale unit cost linearly with data rate ■ Provide meaningful data rate increase ■ Next Ethernet Date Rate products should be introduced in the Parity stage of their underlying technology life cycle ■ Next Ethernet Date Rate products should be based on 100GE technology to control R&D and unit costs 28 May 2010 4 100GE Transceiver Road Map 28 May 2010 5 100GE Gen2 WDM Transceiver 28 May 2010 6 100GE Gen2 MPO Transceiver 28 May 2010 7 400GE WDM Transceiver 28 May 2010 8 400GE WDM Transceiver Specs ■ ITU G.694.1 DWDM grid ■ 800GHz LAN WDM spacing and 1310nm band (same as 100GE) ■ SOA closes the higher 400Gb/s (vs. 100Gb/s) 10km link budget due to: ■ ■ higher fiber loss and dispersion penalty 70nm span (vs. 15nm) ■ higher Mux and DeMux loss 16:1 (vs. 4:1) 400GE-ER16 transceiver can support max reach of ~20km (vs.40km) 28 May 2010 9 400GE MPO Transceiver 28 May 2010 10 1TE Transceivers ■ ■ 100GE Technology based Architecture ■ 40x25G (WDM or MPO) ■ Not practical 1TE technology will require non-NRZ electrical I/O ■ ■ ■ Ex. multi-level amplitude modulation 1TE technology will require non-NRZ optical I/O ■ Ex. 1: multi-level amplitude modulation ■ Ex. 2: phase modulation ■ Ex. 3: optical polarity multiplexing Such technology will not exist for the Ethernet market in the near term and will require major R&D investment 28 May 2010 11 Does 1TE Justify Major R&D? ■ ■ ■ ■ ■ High end data center operator unit interface cost requirement (Google presentations): 10x Δrate → <4x Δcost (mature technology) Resulting 1TE transceiver unit cost: ≈ 4x 100GE cost ≈ 4x (4x 10GE cost) = 16x 10GE-LR SFP+ cost Total potential 1TE port total (Facebook presentations) ■ 1TE link total per high end data center: ~100 ■ High end data center number: ~200 Total potential 1TE market port cost: ~300,000x 10GE-LR SFP+ cost (tiny ROI!) 1TE does not justify major R&D investment at this time 28 May 2010 12 Summary ■ Next Data Rate should be driven by cost requirements ■ Next Data Rate should be based on 100GE technology ■ 400GE Transceivers can be based on 100GE technology ■ 1TE Transceivers can not be based on 100GE technology ■ 1TE does not justify major R&D investment at this time ■ The Next Ethernet Data Rate should be 400Gb/s 28 May 2010 13