Sunday, 05 March,
16:00
–
18:30
Room 8
Description:
There are different approaches to improing energy efficiency in and between data centers. We can continue to press toward higher baud rates and higher bits per symbol or use multiplexing techniques over optical wavelength, subcarrier frequency, and spatial/fiber cores at lower baud rates. Which approach results in the best efficiency in terms of overall power consumption for the end-to-end system? IT should include the power consumption of electrical interconnect, light generation, detection, amplification, signal processing, SERDES, and FEC. This workshop will address this topic in two sessions, within the data center (chip to chip) and between data centers.
Short Reach, <100 meters
The first session will deal with short-reach interconnect within a rack or between racks. Energy per bit is vital in these short optical interconnects.
- In the context of the bandwidth needed for 100 Tbps switches, where is the sweet spot in the battle between high channel bit rates vs. high fiber counts and channels per fiber?
- In lieu of higher channel data rates, what’s the best way to increase the number of channels per fiber core or fiber cores around the host ASIC?
- On the host ASIC side, can we save power with the wider and slower electrical interfaces made possible with co-packaged optics?
Long Reach, up to 10km
The second session will look at communications between data centers.
- Which technologies and multiplexing techniques promise the most energy efficiency to achieve these distances? Is it intensity modulation/direct detection on multiple lanes, coherent subcarrier multiplexing, simplified coherent, or something else?
- What’s the role of DSP in a power-efficient transponder? Does DSP actually result in net power savings?
- What key technologies must be demonstrated to realize these energy-efficient techniques in practical product offerings?
Organizers
Di Che, Nokia Bell Labs, USA
Trey Greer, NVIDIA, USA
Norm Swenson, Norman Swenson Consulting, USA
Xian Zhou, University of Science & Technology Beijing, China
Speakers
Keren Bergmen, Columbia University, USA
Arash Farhood, Marvell Technology, USA
Joe Kahn, Stanford University, USA
Dan Kuchta, IBM, USA
Di Liang, Alibaba Group, USA
Karl Muth, Broadcom, USA
Sunil Priyadarshi, Intel Corporation, USA
Dave Welch, Infinera, USA