Jump to Content
Hong Liu

Hong Liu

Hong Liu is a Google Fellow at Machine Learning, Systems and Cloud AI (MSCA), where she is involved in the system architecture and interconnect for a large-scale computing platform. Her research interests include high speed signaling, optical architecture and interconnection for datacenter networking and machine learning. Prior to joining Google, Hong was a Member of Technical Staff at Juniper Networks, where she worked on the architecture and design of network core routers and multi-chassis switches. She holds 55 U.S. patents in the areas of photonic devices, datacenter networks, and optical communications. Hong received her Ph.D in electrical engineering from Stanford University, is an OSA Fellow.
Authored Publications
Google Publications
Other Publications
Sort By
  • Title
  • Title, descending
  • Year
  • Year, descending
    Preview abstract This is an invited OFC 2024 conference workshop talk regarding a new type of lower-power datacenter optics design choice: linear pluggable optics. In this talk I will discuss the fundamental performance constraints facing linear pluggable optics and their implications on DCN and ML use cases View details
    Preview abstract We review state of art datacenter interconnect technologies for 800G, 1.6T and beyond networking speeds, with a special emphases on 200G per-lane IM-DD and 800G-LR1 coherent-lite View details
    Preview abstract We review state-of-the-art datacenter technologies for 800G, 1.6T and beyond interconnect speeds, focusing on 200G per-lane IM-DD (intensity modulated-direct detect) and 800G-LR1 coherent-lite transmissions. View details
    Preview abstract This is an OFC 2022 conference panel talk. We review from high level about the right technology choice for 1.6Tb/s data center optics interconnects View details
    Preview abstract We discuss technology options and challenges for beyond 1Tb/s intra-datacenter interconnect bandwidth scaling, with a focus on pulse amplitude modulation (PAM)-based intensity modulation-direct detection (IM-DD) and baud-rate digital signal processing (DSP)-based coherent technology. We show that a flexible PAM (FlexPAM) technique with fine granularity of bandwidth/spectral efficiency (in terms of bits/symbol) could bring significant value for 200Gb/s per lane or per dimension bandwidth scaling. For baud-rate sampled coherent technology, we propose a much simplified DSP architecture which could bring down coherent DSP power close to that of IM-DD PAM technology. We also propose two new phase noise tolerant, 5 bits per symbol, 2D coherent modulation formats to lower the requirement on laser linewidth. To conclude, we present a comparative study of IM-DD PAM versus coherent PM-QAM (polarization multiplexed-quadrature amplitude modulation) for 200Gb/s per dimension throughput scaling in terms of achievable link loss/reach budget, power consumption, implementation complexity, as well as fan-out granularity. View details
    Beyond 1Tb/s Datacenter Interconnect Technology: Challenges and Solutions
    OFC 2019 conference proceedings, OSA, The Optical Society 2010 Massachusetts Ave, NW Washington, DC 20036 USA (2019)
    Preview abstract We discuss challenges and solutions for beyond 1Tb/s intra-datacenter bandwidth scaling, with a focus on the FlexPAM-based direct detection and the baud-rate DSP enabled coherent detection. View details
    Preview abstract In this presentation, we will discuss Google’s intra-datacenter networks and interconnect. We will first review the evolution of datacenter interconnects and networking over the past decade, then outline future technology directions which will be needed to keep pace with the requirements and growth of the datacenter. View details
    Preview abstract In this presentation, we will review the evolution of Google’s intra-datacenter interconnects and networking over the past decade, then outline future technology directions which, along with a more holistic design approach, will be needed to keep pace with the requirements and growth of the datacenter. View details
    Preview abstract In this workshop talk we will discuss the potential and value of flexible modulation formats and constellation shaping for datacenter interconnection bandwidth scaling View details
    Preview abstract Increasing demands for web and cloud-based services have been driving exponential growth of datacenter bandwidth. This paper discusses, from Google’s perspective, emerging challenges and possible technical solutions to scale intra-datacenter and intra-campus interconnection network bandwidth. View details
    Preview abstract This paper reviews over decade of technology evolution and advancement of datacenter optical interconnect, which is mainly driven by the explosive bandwidth demand growth of web and cloud based services. Emerging trend and technical options to scale bandwidth well beyond 400Gb/s have also been discussed. View details
    Jupiter Rising: A Decade of Clos Topologies and Centralized Control in Google's Datacenter Network
    Joon Ong
    Amit Agarwal
    Glen Anderson
    Ashby Armistead
    Roy Bannon
    Seb Boving
    Gaurav Desai
    Bob Felderman
    Paulie Germano
    Anand Kanagala
    Jeff Provost
    Jason Simmons
    Eiichi Tanda
    Jim Wanderer
    Stephen Stuart
    Communications of the ACM, vol. Vol. 59, No. 9 (2016), pp. 88-97
    Preview abstract We present our approach for overcoming the cost, operational complexity, and limited scale endemic to datacenter networks a decade ago. Three themes unify the five generations of datacenter networks detailed in this paper. First, multi-stage Clos topologies built from commodity switch silicon can support cost-effective deployment of building-scale networks. Second, much of the general, but complex, decentralized network routing and management protocols supporting arbitrary deployment scenarios were overkill for single-operator, pre-planned datacenter networks. We built a centralized control mechanism based on a global configuration pushed to all datacenter switches. Third, modular hardware design coupled with simple, robust software allowed our design to also support inter-cluster and wide-area networks. Our datacenter networks run at dozens of sites across the planet, scaling in capacity by 100x over 10 years to more than 1 Pbps of bisection bandwidth. View details
    Preview abstract This is an invited OSA annual meeting talk. This talk reviews technology options and challenges to scale bandwidth and power for large-scale data center interconnects. View details
    Pluggable DWDM: Considerations For Campus and Metro DCI Applications
    ECOC 2016 workshop presentation, VDE, Germany, Stresemannallee 15 60596 Frankfurt/Main Germany (2016)
    Preview abstract This is a ECOC workshop talk. We will discuss, from Google’s perspective, the challenges facing datacenter campus and metro interconnection networks, and the potential of using pluggable tunable DWDM for both campus and metro applications View details
    Preview abstract We give an overview of optical interconnect requirements for large scale datacenters. We then make a comparison between silicon photonics technologies and more traditional options in meeting these requirements. View details
    Preview abstract We discuss the trend in fiber optic technology developments to fulfill the scaling requirements of datacenter networks. View details
    Optical Interconnects for Scale-Out Data Centers
    Amin Vahdat
    Optical Interconnects for Future Data Center Networks, Springer, Avenel, NJ (2013), pp. 17-31
    Preview abstract We review the architecture of modern datacenter networks, as well as their scaling challenges; we then present opportunities and needs for emerging optical technologies to support datacenter scaling. View details
    High Performance, Low Cost, Colorless ONU for WDM-PON
    Chris Johnson
    Optical Fiber Communication (OFC) Conference 2012, OSA, Washington DC
    Preview abstract We give an overview of key technologies for realizing WDM-PON. In particular, we highlight promising developments and directions in widely tunable laser technologies for achieving a high performance, colorless ONU at the cost points required for access networks. View details
    Silicon Photonics for Optical Access Networks
    Pedram Dashti
    Chris Johnson
    Group IV Photonics, IEEE, Washington DC (2012), pp. 3
    Preview abstract We highlight promising developments and directions in silicon photonics for realizing cost effective WDM-PON: photonic integration for integrated WDM transceivers at the OLT and widely tunable laser technologies for achieving a high performance, colorless ONU. View details
    The Emerging Optical Data Center
    Amin Vahdat
    Xiaoxue Zhao
    Chris Johnson
    OFC 2011, OTuH2
    Preview abstract We review the architecture of modern datacenter networks, as well as their scaling challenges; then present high-level requirements for deploying optical technologies in datacenters, particularly focusing on optical circuit switching and WDM transceivers. View details
    Hardware Requirements for Optical Circuit Switched Data Center Networks
    Nathan Farrington
    Yeshaiahu Fainman
    George Papen
    Amin Vahdat
    IEEE Communication Society, IEEE Photonic Society, Optical Society of America, OFC/NFOEC 2011, Los Angeles (2010), OTuH3
    Preview abstract Based on measurements of a prototype, we identify hardware requirements for improving the performance of hybrid electrical-packet-switched/optical-circuit-switched data center networks. View details
    Scaling Optical Interconnects in Datacenter Networks Opportunities and Challenges for WDM
    Chris Johnson
    2010 18th IEEE Symposium on High Performance Interconnects, IEEE, pp. 113-116
    Preview abstract We review the growing need for optical interconnect bandwidth in datacenter networks, and the opportunities and challenges for wavelength division multiplexing (WDM) to sustain the “last 2km” bandwidth growth inside datacenter networks. View details
    Fiber Optic Communication Technologies: What’s Needed for Datacenter Network Operations
    Xiaoxue Zhao
    Valey Kamalov
    Vijay Gill
    IEEE Communications Magazine, vol. Vol.48 No.7 (2010)
    Preview abstract The authors review the growing trend of warehouse-scale mega-datacenter computing, the Internet transformation driven by mega-datacenter applications, and the opportunities and challenges for fiber optic communication technologies to support the growth of mega-datacenter computing in the next three to four years. View details
    Energy Proportional Datacenter Networks
    Dennis Abts
    Peter Klausler
    Proceedings of the International Symposium on Computer Architecture, ACM (2010), pp. 338-347
    Preview abstract Numerous studies have shown that datacenter computers rarely operate at full utilization, leading to a number of proposals for creating servers that are energy proportional with respect to the computation that they are performing. In this paper, we show that as servers themselves become more energy proportional, the datacenter network can become a significant fraction (up to 50%) of cluster power. In this paper we propose several ways to design a high-performance datacenter network whose power consumption is more proportional to the amount of traffic it is moving --- that is, we propose energy proportional datacenter networks. We first show that a flattened butterfly topology itself is inherently more power efficient than the other commonly proposed topology for high-performance datacenter networks. We then exploit the characteristics of modern plesiochronous links to adjust their power and performance envelopes dynamically. Using a network simulator, driven by both synthetic workloads and production datacenter traces, we characterize and understand design tradeoffs, and demonstrate an 85% reduction in power --- which approaches the ideal energy-proportionality of the network. Our results also demonstrate two challenges for the designers of future network switches: 1) We show that there is a significant power advantage to having independent control of each unidirectional channel comprising a network link, since many traffic patterns show very asymmetric use, and 2) system designers should work to optimize the high-speed channel designs to be more energy efficient by choosing optimal data rate and equalization technology. Given these assumptions, we demonstrate that energy proportional datacenter communication is indeed possible. View details
    No Results Found