Considering 2km and 10km Architectural Alternatives

Size: px

Start display at page:

Download "Considering 2km and 10km Architectural Alternatives"

氏续冒
5 years ago
Views:

1 Considering 2 km and 10 km Architectural Alternatives Dan Dove, Huawei Xiaolu Song, Huawei Jan. 2014

2 Supporters Francois Tremblay Semtech Sudeep Bhoja Inphi Haijun Wang - China Unicom Xiangkun Man - China Unicom Page 2

3 Outline Defining Broad Market Potential for 2 km and 10 km solutions Market View of Relative Cost for 2 km and 10 km solutions Evaluating Technical Complexity of 2 km and 10 km alternatives Page 3

Broad Market Potential (for 2 km and 10 km solutions) The term Broad Market Potential has changed as Ethernet has evolved from a corporate edge solution into the data center and carrier core solution.

For example, the volume of 400G router-otn connections in the world will be much smaller than the volume of 400G data-center fabrics, but both are necessary to enable a broader market for Ethernet.

4 Broad Market Potential (for 2 km and 10 km solutions) The term Broad Market Potential has changed as Ethernet has evolved from a corporate edge solution into the data center and carrier core solution. Volume for a given technology is less important than where it fits into the market, and how many end users are ultimately served by it. For example, the volume of 400G router-otn connections in the world will be much smaller than the volume of 400G data-center fabrics, but both are necessary to enable a broader market for Ethernet. However, when we further considered what do we really want, the timing question is very relevant, and this question needs to be reworded to what do we really need and when. According to the Lightcounting s presentation at the previous 400GbE ad-hoc meeting, the overview of the need until about 2020, first needs will be in telecom and interconnection between datacenters and telecom. (1) ref: murray_app_01a_1013. Page 4

Broad Market Potential (for 2 km and 10 km solutions) Router (R) OA OTN (O) Inside Building Outside Building Inter Site R IP Router + OTN Metro / LH transmission (>200km) R O R O Outside Building

5 Broad Market Potential (for 2 km and 10 km solutions) Router (R) OA OTN (O) Inside Building Outside Building Inter Site R IP Router + OTN Metro / LH transmission (>200km) R O R O Outside Building Different Sites O 500m~2km ~2km ~10km Metro Network 10km~40km interconnection (even 80km) ~70% of router-otn connections will be served by a 2 km fiber PMD solution. (1) ~15% of router-otn connections cannot be served by a 2 km fiber PMD but can be served by a 10 km PMD. (1) Based upon the current 10GbE situations, 10km can cover at least 60% of link requirements and 40km can cover 80%~90% of requirements. (1) Undoubtedly this application enables a broader market for Ethernet. Given the ratio of 2 km to 10 km market size, the relative cost of these two targets must be carefully considered. (1) ref: song_x_400_01a_1113. Page 5

6 Relative Cost (for 2 km and 10 km solution) Relative Cost requirements at the core are different than they are as they move toward the edge. For longer links, higher cost and complexity may be tolerated in the transceiver if it helps to achieve an overall reduction in system cost. Per-port costs are often amortized over large numbers of users and thus a generic 10x the performance for 3x the cost expectation does not necessarily make sense. Comparing relative costs in the same application for given technology steps is a more practical approach. Consideration of installed vs. new cabling is important because the cost of installation can be substantial in longer reach applications. One size fits all solutions (i.e.: 10 km) can address short-term market needs, but long-term cost focus becomes more important for shorter reach applications. Page 6

7 Relative Cost (for 2 km and 10 km solution) Slide showing relative Cost per Gb actual and projected for various Ethernet transceiver technologies provided by Dale Murray, LightCounting A number of useful observations can be made from this slide. * Ref: Dale Murray - LightCounting Page 7

8 Relative Cost (for 2 km and 10 km solution) Asymptotic Cost for Long Reach (10km) vs. Short Reach (<300m) shows that market can tolerate > 2x Cost per Gb Δ based on reach / application. Relative Cost Asymptote Long Reach (10km) Short Reach (<300m) * Ref: Dale Murray - LightCounting Page 8

9 Relative Cost (for 2 km and 10 km solution) Asymptotic Cost for Long Reach (10km) vs. Short Reach (<300m) shows that market can tolerate > 2x Cost per Gb Δ based on reach / application. Observe that as market matures, it is expected that Cost per Gb should be lower than prior speed. Relative Cost Asymptote Long Reach (10km) Short Reach (<300m) * Ref: Dale Murray - LightCounting Page 9

10 Relative Cost (for 2 km and 10 km solution) Asymptotic Cost for Long Reach (10km) vs. Short Reach (<300m) shows that market can tolerate > 2x Cost per Gb Δ based on reach / application. Observe that as market matures, it is expected that Cost per Gb should be lower than prior speed. Observe that Cost per Gb actually increases for mature technologies as their use / volume is overtaken by newer and faster technologies. Relative Cost Asymptote Long Reach (10km) Short Reach (<300m) * Ref: Dale Murray - LightCounting Page 10

11 Relative Cost Targets (by solution) Table of relative cost targets as a function of reach/application (Cost/Gb). Market DC Edge 3m~100m DC Fabric 100m~500m DC Core 2 km Carrier IP-Core 10 km 400GbE (projected) 0.8t 1.2t 1.5t 2t The value 1t is approximately equal to cost of 100G DC Short Reach solution. Different applications require different cost points to achieve successful market adoption. * Note These are simplified extrapolations, not intended to be exact. Page 11

12 Relative Cost (likely technical alternatives - 2 km and 10 km) Lets explore song_400_01_0513 options 16x25 GBd NRZ 8x50 GBd NRZ 8x25 GBd PAM4 4x50 GBd PAM4 4x100Gb/s HOM (>PAM4) Page 12

13 Technical Complexity (2 km and 10 km) 16x25 GBd NRZ Laser & modulation straight-forward / low cost per unit but number of units add cost Tighter wavelength control required - adds cost Or expanded grid (2~4*LWDM) increases dispersion penalty Mux complexity and filter complexity higher - adds cost 16-channel Mux and Demux increase loss (PLC platform) More components, more power, more complex packaging adds cost 400GbE LR16 Module CDR etc x16 25 GBd Drivers x16 LD +EAM (x16) x16 Mux Duplex SMF CDR/EQ etc x16 25 GBd TIAs x16 PD (x16) x16 DeMux SOA Addition of SOA or APD may be required for 10km. Page 13

14 Technical Complexity (2 km and 10 km) 8x50 GBd NRZ Lasers & modulation more complex, higher cost per unit but reduced number of units cost TBD Tight wavelength control required - adds cost Or expanded grid (2*LWDM) increases dispersion penalty Mux complexity and filter complexity high - adds cost Many components, high power, more complex packaging adds cost 400GbE LR8 Module CDR etc x8 50 GBd Drivers x8 LD+ EAM (x8) x8 Mux Duplex SMF CDR/EQ etc x8 50 GBd TIAs x8 PD (x8) x8 DeMux SOA Addition of SOA or APD may be required for 10km. Page 14

15 Technical Complexity (2 km and 10 km) 8x25 GBd PAM4 Commoditized Lasers, modulation more complex, balanced cost per unit but reduced number of units cost TBD Tight wavelength control required - adds cost Or wider grid (2xLWDM) adds dispersion penalty Mux complexity and filter complexity high - adds cost Many components, high power, more complex packaging adds cost Balances Time-To-Market with Cost 400GbE LP8 Module PAM4 Encoder x8 25GBd Linear Drivers x8 LD+ EAM (x8) x8 Mux Duplex SMF PAM4 Decoder CDR/EQ x8 25GBd Linear TIAs x8 PD (x8) x8 DeMux SOA Addition of SOA or APD may be required for 10km. Page 15

16 Technical Complexity (2 km only?) 4x50 GBd PAM4 Lasers & modulation more complex, higher cost per unit but substantially reduced number of units saves cost Less wavelength control required - saves cost Mux complexity and filter complexity lower - saves cost Fewer components, lower power, simpler packaging saves cost Time to market later, but ultimately lowest cost 400GbE LP4 Module PAM4 Encoder x4 50 GBd Linear Drivers x4 Tx (x4) x4 Mux Duplex SMF PAM4 Decoder CDR/EQ x4 50 GBd Linear TIAs x4 Rx (x4) x4 DeMux Page 16

17 Technical Complexity (2 km or 10 km) 4x100Gb/s HOM (>PAM4) Lasers & modulation more complex, higher cost per unit but substantially reduced number of units saves cost Less wavelength control required saves cost Mux complexity and filter complexity lower saves cost Fewer optical components, lower power, but bigger digital components and power may challenge packaging Time to market late, complexity needs to be understood 400GbE HOM4 Module HOM Encoder x4 xxgbd Linear Drivers x4 Tx (x4) x4 Mux Duplex SMF HOM Decoder CDR/EQ x4 xxgbd Linear TIAs x4 Rx (x4) x4 DeMux Page 17

18 Technical Complexity & Cost Balance (2 km) Solutions 4x25 GBd NRZ 16x25 GBd NRZ 8x25 GBd PAM4 8x50 GBd NRZ 4x50 GBd PAM4 Relative Cost of TOSA Relative Cost of ROSA Relative Cost of IC Relative Total Cost x25 GBd PAM4 solution achieves Technical Complexity (Time-To- Market) balance with lowest cost. 4x50 GBd PAM4 solution achieves lowest cost in the longer term. Page 18

19 Technical Complexity & Cost Balance (10 km) Solutions 4x25 GBd NRZ 16x25 GBd NRZ 8x25 GBd PAM4 8x50 GBd NRZ 4x50 GBd PAM4 Relative Cost of TOSA Relative Cost of ROSA Relative Cost of IC Relative Total Cost ?? ?? ?? ?? 8x25 GBd PAM4 solution achieves Technical Complexity (Time-To- Market) balance with lowest cost. 4x50 GBd PAM4 solution not proven via simulation to work for 10 km Page 19

20 Technical Complexity & Cost Balance Cost 16x 25 GBd NRZ Power Technical Complexity While relatively low in technical complexity, a high number of channels leads to higher cost, power and lower reliability. Page 20

21 Technical Complexity & Cost Balance Cost 8x50 GBd NRZ Power Technical Complexity While reducing the number of channels can reduce cost, the complexity of those channels goes up. 50 GBd NRZ devices are not yet readily available and will push the envelope of technology. Ultimately, this approach should lower power. Page 21

22 Technical Complexity & Cost Balance Cost 8x25 GBd PAM4 Power Technical Complexity By reducing the number of channels, reduces cost of optics. By using lower baud rate, is more technically available and lower cost components can be used. Power is lower than 16x25 GBd NRZ, and is more reliable Page 22

23 Technical Complexity & Cost Balance Cost 4x50 GBd PAM4 Power Technical Complexity By reducing the number of channels down to 4, substantially reduces cost of optics. Higher baud rate, is more technically complex and but long-term lower cost solution. The power is lower than 8x25 GBd PAM4, and is more reliable Page 23

24 Technical Complexity & Cost Balance Cost >PAM4 HOM Power Technical Complexity Due to technical complexity and power challenges, >PAM4 HOM remains unrealistic for this generation of standard. Longer term, offers a future for higher speed and lowest cost. Page 24

25 Technical Complexity & Cost Balance Cost 16x25 GBd NRZ 8x50 GBd NRZ 8x25 GBd PAM4 4x50 GBd PAM4 >PAM4 HOM Power Technical Complexity 8x25 GBd PAM4 solution achieves Technical Complexity (Time-To- Market) balance with lowest cost in the shorter term. 4x50 GBd PAM4 solution achieves lowest cost in a practical timeframe. Page 25

26 Conclusions ( 2 km and 10 km Objectives) Market demand for 2 km and 10 km solution have been demonstrated throughout Study Group effort (BMP) Cost requirement for the 10 km solution is ~1.7x the cost of a 2 km solution and both can come in below 100G Cost per Gb for comparable application/ solutions (EF) Technical approaches are considered, and multiple solutions appear feasible and worthy of consideration (TF) (EF) Time-to-market for 8x25 GBd PAM4 is slightly lower than 4x50 GBd PAM4 (lowest cost alternative) We believe the market can tolerate that higher cost in the short term and migrate to 4x50 GBd PAM4 solution as market matures Recommendation: Continue to investigate and refine research on 10 km and 2 km proposals Page 26

PowerPoint Presentation

PowerPoint Presentation Current Status of the 10G-EPON Power Budget and Improvement David Li, Hisense-Ligent dli@ligentphotonics.com Ligent Supporters Bo Wang, China Telecom David Li, Ligent Photonics Frank Chang, Vitesse Corp.