{"id":679395,"date":"2020-08-21T08:15:31","date_gmt":"2020-08-21T15:15:31","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-group&#038;p=679395"},"modified":"2024-03-06T09:36:02","modified_gmt":"2024-03-06T17:36:02","slug":"optics-for-the-cloud-research-alliance","status":"publish","type":"msr-group","link":"https:\/\/www.microsoft.com\/en-us\/research\/collaboration\/optics-for-the-cloud-research-alliance\/","title":{"rendered":"Optics for the Cloud Research Alliance"},"content":{"rendered":"<section class=\"mb-3 moray-highlight\">\n\t<div class=\"card-img-overlay mx-lg-0\">\n\t\t<div class=\"card-background  has-background- card-background--full-bleed\">\n\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"1920\" height=\"720\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header.jpg\" class=\"attachment-full size-full\" alt=\"illustration of clouds within a cloud connected by data points with a rainbow of colored lines shooting downward\" style=\"object-position: 73% 51%\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header.jpg 1920w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header-300x113.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header-1024x384.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header-768x288.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header-1536x576.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/08\/optics-for-the-cloud-header-1600x600.jpg 1600w\" sizes=\"(max-width: 1920px) 100vw, 1920px\" \/>\t\t<\/div>\n\t\t<!-- Foreground -->\n\t\t<div class=\"card-foreground d-flex mt-md-n5 my-lg-5 px-g px-lg-0\">\n\t\t\t<!-- Container -->\n\t\t\t<div class=\"container d-flex mt-md-n5 my-lg-5 align-self-center\">\n\t\t\t\t<!-- Card wrapper -->\n\t\t\t\t<div class=\"w-100 w-lg-col-5\">\n\t\t\t\t\t<!-- Card -->\n\t\t\t\t\t<div class=\"card material-md-card py-5 px-md-5\">\n\t\t\t\t\t\t<div class=\"card-body \">\n\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\n\n<h1 class=\"wp-block-heading h2\" id=\"optics-for-the-cloud-research-alliance\">Optics for the Cloud Research Alliance<\/h1>\n\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n\n\n\n\n\n<p>Microsoft Research Cambridge has brought together its Optics for the Cloud collaborations with several university research groups in Europe and the US to create the Optics for the Cloud Research Alliance. This Alliance investigates and develops optical technologies for the cloud. The Alliance takes a multi-disciplinary approach to tackle big and open challenges, with an aim to enable greater adoption of optical technologies in cloud infrastructure, including cloud storage, network, and compute.<\/p>\n\n\n\n<p>The research spans a significant part of the cloud infrastructure ecosystem, ranging from photonics fabrication, via optical system and sub-system design, to co-design and integration with the rest of the data centre stack. The Alliance provides a vibrant collaborative research environment, involving experts across optical and cloud research disciplines, with an open publications policy. The Alliance funds a cohort of PhD scholars. The three-year PhDs target long-term problems and aim to advance the state of the art through forward-looking fundamental research that is being published in top technical venues. Each PhD student is hosted and supervised by a university faculty member and co-supervised by a Microsoft researcher. The students also benefit from discussion, insights, and feedback from practitioners at Microsoft and, more broadly, through greater interaction with partners in industry and academia.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"goals-of-the-research-alliance\">Goals of the Research Alliance<\/h3>\n\n\n\n<p>The primary goals of the alliance are to:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Leverage and enable true cross-disciplinary research<\/li>\n\n\n\n<li>Support high-impact pre-competitive research<\/li>\n\n\n\n<li>Aspire to generate disruptive, not sustaining, ideas<\/li>\n<\/ul>\n\n\n\n<p>The secondary goals are to:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Help to expand the ecosystem<\/li>\n\n\n\n<li>Communicate the value of optics and photonics research in UK and Europe<\/li>\n\n\n\n<li>Send a message that we view our members as world leaders in this space<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"members-of-the-research-alliance\">Members of the Research Alliance<\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-1 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.tue.nl\/en\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/07\/TUE_logo_300x200.png\" alt=\"Eindhoven University of Technology logo\" class=\"wp-image-599781\"\/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.imperial.ac.uk\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Imperial_College_London_logo_300x200.jpg\" alt=\"Imperial College London logo\" class=\"wp-image-710416\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Imperial_College_London_logo_300x200.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Imperial_College_London_logo_300x200-16x12.jpg 16w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.epfl.ch\/en\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/09\/EPFL_300x200.jpg\" alt=\"Swiss Federal Institute of Technology Lausanne (EPFL) logo\" class=\"wp-image-691188\"\/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-2 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucsb.edu\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/07\/UnivSantaBarbara_logo_300x200.png\" alt=\"University of California Santa Barbara logo\" class=\"wp-image-599790\"\/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.cam.ac.uk\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/07\/UnivCambridge_logo_300x200.png\" alt=\"University of Cambridge logo\" class=\"wp-image-599787\"\/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/04\/UCL_logo_300x200.png\" alt=\"University College London (UCL) logo\" class=\"wp-image-738370\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/04\/UCL_logo_300x200.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/04\/UCL_logo_300x200-16x12.png 16w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-3 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ox.ac.uk\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Oxford_logo_300x200.jpg\" alt=\"University of Oxford logo\" class=\"wp-image-710422\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Oxford_logo_300x200.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/12\/Oxford_logo_300x200-16x12.jpg 16w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.southampton.ac.uk\/\" target=\"_blank\" rel=\"noopener\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"200\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2018\/07\/UnivSouthampton_marine_logo_300x200.png\" alt=\"University of Southampton logo\" class=\"wp-image-599793\"\/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><\/div>\n<\/div>\n\n\n\n\n\n\n\n<p><\/p>\n\n\n\n\n\n\n\n<p><strong>TU\/e PIs:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.tue.nl\/en\/research\/researchers\/xaveer-leijtens\/\" target=\"_blank\" rel=\"noreferrer noopener\">Xaveer Leijtens<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.tue.nl\/en\/research\/researchers\/kevin-williams\/\" target=\"_blank\" rel=\"noreferrer noopener\">Kevin Williams<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/kaishi\/\">Kai Shi<\/a><\/p>\n\n\n\n<p><strong>TU\/e Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/nl.linkedin.com\/in\/wjtian\" target=\"_blank\" rel=\"noreferrer noopener\">Wenjing Tian<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>The exponential rise in data consumption in the cloud is now pushing data centers beyond their limits. Optical interconnect can be an alternative technology to boost the ever-evolving cloud. Optical switching using photonic integrated circuits holds the promise of flexibility, low latency, connectivity, and energy efficiency. However, the assembly of integrated photonic switches is one of the greatest barriers to deploying optical switch technology. In this work, we investigate the assembly method for photonic integrated circuits using hybrid integration across two photonic integration platforms, Indium Phosphide (InP), and Silicon Nitride (SiN) based TriPleX. The assembly method is suited to flip-chip attachment which offers a cost-effective solution for low loss and high port count fiber-to-chip coupling.<\/p>\n\n\n\n\n\n<p><strong>EPFL PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/people.epfl.ch\/tobias.kippenberg\" target=\"_blank\" rel=\"noreferrer noopener\">Tobias Kippenberg<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/sophie-lange-9791a227\/\" target=\"_blank\" rel=\"noreferrer noopener\">Sophie Lange<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>EPFL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/people.epfl.ch\/arslan.raja\/\" target=\"_blank\" rel=\"noreferrer noopener\">Arslan Raja<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>The sub-nanoseconds (less than 550 ps) optical switching is achieved by using a multiwavelength source based on a soliton microcomb (Si3N4) and wavelength selector based on a semiconductor optical amplifier. The optical switching systems with 25 GBd NRZ and PAM4 burst mode transmission are demonstrated.<\/p>\n\n\n\n\n\n<p><strong>ICL PIs:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.imperial.ac.uk\/people\/mark.neil\" target=\"_blank\" rel=\"noreferrer noopener\">Mark Neil<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.imperial.ac.uk\/people\/matthew.foreman\" target=\"_blank\" rel=\"noreferrer noopener\">Matthew Foreman<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jaclegg\/\">James Clegg<\/a><\/p>\n\n\n\n<p><strong>ICL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/feng-zhonghe-2a127713a\/.\">Zhonghe Feng<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>Measuring birefringence in 3D is important for many applications including optical data storage and cell biology. The current state-of-the-art techniques use methods that are designed to measure thin (2D) structures but have been adapted in ad-hoc ways and combined with advanced processing to recover 3D information. Remarkably, there is an important and basic research question that still exists: is there a physically correct way to isolate and measure the birefringence of a thin 2D layer that is contained within a 3D volume, and therefore correctly reconstruct a full 3D volume? This PhD research project will answer this question.<\/p>\n\n\n\n\n\n<p><strong>UCL PIs:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/prof-polina-bayvel\">Polina Bayvel<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/iccs\/prof-robert-killey\" target=\"_blank\" rel=\"noreferrer noopener\">Robert Killey<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\">George Zervas<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PIs:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/chrisgk\/\">Christos Gkantsidis<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fokarino\/\">Fotini Karinou<\/a><\/p>\n\n\n\n<p><strong>UCL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/hubertdzieciol\/\" target=\"_blank\" rel=\"noreferrer noopener\">Hubert Dzieciol<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>Signal design strategies to improve phase noise tolerance of a short-range coherent system such as 400ZR and beyond. We propose a low-complexity approach demonstrating up to .7dB gains (both GMI and post-FEC) when compared to 16QAM constellations and 0.4dB for AWGN optimal formats.<\/p>\n\n\n\n\n\n<p><strong>UCL PIs:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/prof-polina-bayvel\" target=\"_blank\" rel=\"noreferrer noopener\">Polina Bayvel<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>,\u202f<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\" target=\"_blank\" rel=\"noreferrer noopener\">George Zervas<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/bethomse\/\">Benn Thomsen<\/a><\/p>\n\n\n\n<p><strong>UCL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/thomas-gerard-9390a05b\/?originalSubdomain=uk\" target=\"_blank\" rel=\"noreferrer noopener\">Thomas Gerard<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>The aim of this PhD project was to develop an intelligent transmitter that can enable the creation of all-optical architectures for flat, scalable, high bandwidth data centre networks. The transmitter must exceed the capabilities of existing designs, providing 2 orders of magnitude improvement in switch time (sub-nanosecond), scalable tuning bandwidths covering the C-band and beyond, and power efficient modulation for high link-budget networks. To that end, the thesis investigated four main areas: (1) Determine the transceiver characteristics needed to create scalable wavelength routed networks supporting >105 servers using real time, space- and wavelength-switched optical transceivers operating at 25Gb\/s. (2) Investigation of artificial intelligent methods for commercial tuneable laser optimisation, guaranteeing fast and reliable wavelength switching. (3) Development of novel wavelength tuneable subsystems that can approach and overcome the tuning speeds and bandwidth limits of traditional tuneable laser sources, achieving sub-ns switching across greater-than-C-band frequencies. (4) Investigation of novel transmitter architectures that maximize link loss budget while providing spectrally efficient modulation for data rates of 100Gb\/s\/wavelength.<\/p>\n\n\n\n\n\n<p><strong>UCL PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\" target=\"_blank\" rel=\"noreferrer noopener\">Georgios Zervas<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/pcosta\/\">Paolo Costa<\/a><\/p>\n\n\n\n<p><strong>UCL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/alessandro-ottino-09ba1a14b\/\" target=\"_blank\" rel=\"noreferrer noopener\">Alessandro Ottino<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>In the last few years the demand for Machine Learning (ML) and Artificial Intelligence (AI) has experienced an unprecedented growth due to the improvement and availability of hardware accelerators. To increase the prediction capabilities and applications of deep learning models, deeper and larger models trained with larger amount of data are needed. The requirements of these new models is outpacing the hardware improvement, forcing scientists to rely on distributing the ML workload onto multiple accelerators instead of a single one. Scaling these distributed-training workloads is often limited by communication\/ computation time imbalance, making the network one of the key bottlenecks. The goal of this project is to investigate how novel optics technologies could address these challenges by rethinking the network from the grounds up and purposely optimizing it for distributed training.<\/p>\n\n\n\n\n\n<p><strong>UCL PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.zhixinliu.net\/\" target=\"_blank\" rel=\"noreferrer noopener\">Zhixin Liu<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PIs:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/hiballan\/\">Hitesh Ballani<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/pcosta\/\">Paolo Costa<\/a><\/p>\n\n\n\n<p><strong>UCL Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ucl.ac.uk\/iccs\/kari-clark\" target=\"_blank\" rel=\"noreferrer noopener\">Kari Clark<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>The rapid growth in the amount of data being transferred within data centres, combined with the slowdown in Moore\u2019s Law, creates challenges for the future scalability of electronically switched data-centre networks. Optical switches could offer a future-proof alternative, and photonic integration platforms have been demonstrated with nanosecond-scale optical switching times. End-to-end switching time is, however, currently limited by the clock and data recovery time, which typically takes microseconds, removing the benefits of nanosecond optical switching. Here we show that a clock phase caching technique can provide clock and data recovery times of under 625\u2009ps (16 symbols at 25.6\u2009Gb\u2009s\u22121). Our approach uses the measurement and storage of clock phase values in a synchronized network to simplify clock and data recovery versus conventional asynchronous approaches. We demonstrate the capabilities of our technique using a real-time prototype with commercial transceivers and validate its resilience against temperature variation and clock jitter.<\/p>\n\n\n\n\n\n<p><strong>UCSB PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/iee.ucsb.edu\/people\/daniel-blumenthal\" target=\"_blank\" rel=\"noreferrer noopener\">Dan Blumenthal<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/hiballan\/\">Hitesh Ballani<\/a><\/p>\n\n\n\n<p><strong>UCSB Students:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/grant-brodnik-42b807a1\/\" target=\"_blank\" rel=\"noreferrer noopener\">Grant Brodnik<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/ocpi.ece.ucsb.edu\/people-2\/\" target=\"_blank\" rel=\"noreferrer noopener\">Mark Harrington<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>As data center ethernet switches scale toward 100 Tbps, the fiber interconnect will face significant cost, power, and engineering barriers. We describe FRESCO, a fiber data center interconnect (DCI) that brings high order coherent WDM Terabit links inside the data center without the need for DSP and other power consuming technologies. FRESCO replaces high power DSP- and PLL-based carrier phase recovery by employing integrated ultra-low linewidth stimulated Brillouin scattering (SBS) 1550nm lasers stabilized to reference \u00b5-cavities for ~1Hz fundamental linewidth and 10s of Hz integral linewidth optical carriers. This enables the optical frequency stabilized phase locked loop (OFS-PLL) using low bandwidth (<1MHz), low power simple electronics to achieve a direct optical carrier lock with residual phase error variance of 3&#215;10-4 rad2, sufficient for high-order QAM. The architecture is extended to shared source WDM operation using Kerr soliton optical frequency comb generators, amortizing cost and complexity over all channels. Fully integrated silicon photonic transceivers offer reduced footprint and power consumption.<\/p>\n\n\n\n<p>To date, we have demonstrated a discrete component single channel 50Gbd 16-QAM FRESCO link using the DSP-free OFS-PLL with low bandwidth (<1MHz), low power electronics for carrier phase recovery over a 200m fiber channel achieving EVM performance within ~1% of a DSP based reference link. We have also demonstrated generation of an ultra-low linewidth WDM shared source using the frequency stabilized SBS laser to pump a Kerr soliton in a SiN microring. Operation as a shared source architecture offers unique opportunities for power savings and motivates current efforts in modeling the impact of stable, ultra-low phase noise optical sources on link performance and power consumption. With our proposed DSP-free architecture, demands on data-chain electronics such as DACs and ADCs may be relaxed, offering additional power savings for tomorrow\u2019s hyperscale DCIs. Future experimental efforts will scale baudrate, QAM order, and WDM count approaching energy efficiency below 10pJ \/bit and 10s of Terabits link capacity.<\/p>\n\n\n\n\n\n<p><strong>CU PIs:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/rvp11\" target=\"_blank\" rel=\"noreferrer noopener\">Richard Penty<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www-g.eng.cam.ac.uk\/photonic_comms\/files\/cps_groupmembers_page.php?crsid=nb301\" target=\"_blank\" rel=\"noreferrer noopener\">Nikos Bamiedakis<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fokarino\/\">Fotini Karinou<\/a><\/p>\n\n\n\n<p><strong>CU Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.linkedin.com\/in\/arastu-sharma-13564ab2\/\" target=\"_blank\" rel=\"noreferrer noopener\">Arastu Sharma<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>Incorporating optics in close proximity to high-performance electronic chips on the board-level can not only allow substitution of the existing electrical connections with faster, more power-efficient alternatives, but most importantly, the radical re-design of the system architecture and yet has had limited exploration. Performance of optical links have interdependence on the link utilisation by the application traffic, which makes optical design space a multi-dimensional challenge. We need a solution that looks at the problem from different perspectives, and considers all aspects including the allocation of system resources, technologies and network topology for different applications. Through our study we want to develop a tool that can help in quantifying these design trade-offs, which will in turn help us in designing optimal architectures functioning with high bandwidth, energy efficient optical links. In this project therefore, we study various board-level interconnection architectures for multiple computing chips using on-board optics taking into account recent achievements in the field. To conduct a performance metric simulation, we input our designs into an architecture simulator in combination of an opto-electronic device modelling tool and run for standard workload benchmarks. We compare our design trade-offs with respect to each architecture based on the different available technologies and derive the optimum configuration for execution time, power consumption and implementation complexity.<\/p>\n\n\n\n\n\n<p><strong>CU PI:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/dpc31\" target=\"_blank\" rel=\"noreferrer noopener\">Daping Chu<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Microsoft PI: <\/strong><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jaclegg\/\">James Clegg<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/angeor\/\">Andreas Georgiou<\/a><\/p>\n\n\n\n<p><strong>CU Student:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/jh2101\" target=\"_blank\" rel=\"noreferrer noopener\">Jintao Hong<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>A low-cost Liquid Crystal display can modulate the polarization of more than 2 billion of pixels per second. These devices are everywhere, from our smartphones, digital cameras and smart watches and provide the highest electro-optical modulation bandwidth in consumer products. When used in a birefringence-based optical data storage system (like Project Silica), they can enable highly parallel data writing making opto-mechanics simpler and relaxing the femtosecond laser requirements. Nevertheless, there are still many challenges including (a) splitting the laser beam and controlling the polarization of each beam (b) controlling the amplitude of the un-modulated light and (c) ensuring the multiple beams are all tightly focused in the glass.<\/p>\n\n\n\n\n\n<p><strong>UO PI:<\/strong> Harish Bhaskaran<\/p>\n\n\n\n<p><strong>Microsoft PI:<\/strong> Francesca Parmigiani<\/p>\n\n\n\n<p>The use of functional materials that can accumulate information to carry out both memory and computing tasks in-situ is a growing field. Using optical integration onto silicon chips provides a unique opportunity to combine the benefits of silicon scaling and the wavelength multiplexing of optics onto a single platform. In this proposal, we will significantly expand our initial work on carrying out non-von Neumann computations (such as Vector-Matrix Multiplication directly in hardware) to a larger matrix to prove a lab scale demonstration of the potential for such hardware in real-world computing tasks. In addition, this proposal will aim to benchmark such computations against existing state-of-the-art in terms of energy and operational speed.<\/p>\n\n\n\n\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Microsoft Research Cambridge has brought together its Optics for the Cloud collaborations with several university research groups in Europe and the US to create the Optics for the Cloud Research Alliance. This Alliance investigates and develops optical technologies for the cloud. The Alliance takes a multi-disciplinary approach to tackle big and open challenges, with an aim to enable greater adoption of optical technologies in cloud infrastructure, including cloud storage, network and compute.<\/p>\n","protected":false},"featured_media":682338,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":true,"_classifai_error":"","msr_group_start":"","footnotes":""},"research-area":[13547],"msr-group-type":[243721],"msr-locale":[268875],"msr-impact-theme":[],"class_list":["post-679395","msr-group","type-msr-group","status-publish","has-post-thumbnail","hentry","msr-research-area-systems-and-networking","msr-group-type-collaboration","msr-locale-en_us"],"msr_group_start":"","msr_detailed_description":"","msr_further_details":"","msr_hero_images":[],"msr_research_lab":[199561],"related-researchers":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-projects":[],"related-events":[],"related-opportunities":[],"related-posts":[],"tab-content":[{"id":0,"name":"Projects","content":"[accordion]\r\n[panel header=\"Eindhoven University of Technology | Cost-Effective Assembly for Hybrid Integrated Photonic Switches\"]\r\n<p style=\"margin-bottom: 0px\"><strong>TU\/e PIs:<\/strong> <a href=\"https:\/\/www.tue.nl\/en\/research\/researchers\/xaveer-leijtens\/\" target=\"_blank\" rel=\"noopener\">Xaveer Leijtens<\/a>, <a href=\"https:\/\/www.tue.nl\/en\/research\/researchers\/kevin-williams\/\" target=\"_blank\" rel=\"noopener\">Kevin Williams<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/kaishi\/\">Kai Shi<\/a><\/p>\r\n<strong>TU\/e Student:<\/strong> <a href=\"https:\/\/nl.linkedin.com\/in\/wjtian\" target=\"_blank\" rel=\"noopener\">Wenjing Tian<\/a>\r\n\r\nThe exponential rise in data consumption in the cloud is now pushing data centers beyond their limits. Optical interconnect can be an alternative technology to boost the ever-evolving cloud. Optical switching using photonic integrated circuits holds the promise of flexibility, low latency, connectivity, and energy efficiency. However, the assembly of integrated photonic switches is one of the greatest barriers to deploying optical switch technology. In this work, we investigate the assembly method for photonic integrated circuits using hybrid integration across two photonic integration platforms, Indium Phosphide (InP), and Silicon Nitride (SiN) based TriPleX. The assembly method is suited to flip-chip attachment which offers a cost-effective solution for low loss and high port count fiber-to-chip coupling.\r\n[\/panel]\r\n[panel header=\"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne | Photonic Integrated Multi-Wavelength Sources for Data Centers\"]\r\n<p style=\"margin-bottom: 0px\"><strong>EPFL PI:<\/strong> <a href=\"https:\/\/people.epfl.ch\/tobias.kippenberg\" target=\"_blank\" rel=\"noopener\">Tobias Kippenberg<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/sophie-lange-9791a227\/\" target=\"_blank\" rel=\"noopener\">Sophie Lange<\/a><\/p>\r\n<strong>EPFL Student:<\/strong> <a href=\"https:\/\/people.epfl.ch\/arslan.raja\/\" target=\"_blank\" rel=\"noopener\">Arslan Raja<\/a>\r\n\r\nThe sub-nanoseconds (less than 550 ps) optical switching is achieved by using a multiwavelength source based on a soliton microcomb (Si3N4) and wavelength selector based on a semiconductor optical amplifier. The optical switching systems with 25 GBd NRZ and PAM4 burst mode transmission are demonstrated.\r\n[\/panel]\r\n[panel header=\"Imperial College London | Measurement of Birefringence of 3D Structures\"]\r\n<p style=\"margin-bottom: 0px\"><strong>ICL PIs:<\/strong> <a href=\"https:\/\/www.imperial.ac.uk\/people\/mark.neil\" target=\"_blank\" rel=\"noopener\">Mark Neil<\/a>, <a href=\"https:\/\/www.imperial.ac.uk\/people\/matthew.foreman\" target=\"_blank\" rel=\"noopener\">Matthew Foreman<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jaclegg\/\">James Clegg<\/a><\/p>\r\n<strong>ICL Student:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/feng-zhonghe-2a127713a\/.\">Zhonghe Feng<\/a>\r\n\r\nMeasuring birefringence in 3D is important for many applications including optical data storage and cell biology. The current state-of-the-art techniques use methods that are designed to measure thin (2D) structures but have been adapted in ad-hoc ways and combined with advanced processing to recover 3D information. Remarkably, there is an important and basic research question that still exists: is there a physically correct way to isolate and measure the birefringence of a thin 2D layer that is contained within a 3D volume, and therefore correctly reconstruct a full 3D volume? This PhD research project will answer this question.\r\n[\/panel]\r\n[panel header=\"University College London | Extending Phase Noise Tolerance in UDWDM Access Networks\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UCL PIs:<\/strong> <a href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/prof-polina-bayvel\">Polina Bayvel<\/a>, <a href=\"https:\/\/www.ucl.ac.uk\/iccs\/prof-robert-killey\" target=\"_blank\" rel=\"noopener\">Robert Killey<\/a>, <a href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\">George Zervas<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PIs:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/chrisgk\/\">Christos Gkantsidis<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fokarino\/\">Fotini Karinou<\/a><\/p>\r\n<strong>UCL Student:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/hubertdzieciol\/\" target=\"_blank\" rel=\"noopener\">Hubert Dzieciol<\/a>\r\n\r\nSignal design strategies to improve phase noise tolerance of a short-range coherent system such as 400ZR and beyond. We propose a low-complexity approach demonstrating up to .7dB gains (both GMI and post-FEC) when compared to 16QAM constellations and 0.4dB for AWGN optimal formats.\r\n[\/panel]\r\n\r\n[panel header=\"University College London | Optical Switching for Scalable Data Centre Networks\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UCL PIs:<\/strong> <a href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/prof-polina-bayvel\" target=\"_blank\" rel=\"noopener\">Polina Bayvel<\/a>,\u202f<a href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\" target=\"_blank\" rel=\"noopener\">George Zervas<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/bethomse\/\">Benn Thomsen<\/a><\/p>\r\n<strong>UCL Student:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/thomas-gerard-9390a05b\/?originalSubdomain=uk\" target=\"_blank\" rel=\"noopener\">Thomas Gerard<\/a>\r\n\r\nThe aim of this PhD project was to develop an intelligent transmitter that can enable the creation of all-optical architectures for flat, scalable, high bandwidth data centre networks. The transmitter must exceed the capabilities of existing designs, providing 2 orders of magnitude improvement in switch time (sub-nanosecond), scalable tuning bandwidths covering the C-band and beyond, and power efficient modulation for high link-budget networks. To that end, the thesis investigated four main areas: (1) Determine the transceiver characteristics needed to create scalable wavelength routed networks supporting &gt;105 servers using real time, space- and wavelength-switched optical transceivers operating at 25Gb\/s. (2) Investigation of artificial intelligent methods for commercial tuneable laser optimisation, guaranteeing fast and reliable wavelength switching. (3) Development of novel wavelength tuneable subsystems that can approach and overcome the tuning speeds and bandwidth limits of traditional tuneable laser sources, achieving sub-ns switching across greater-than-C-band frequencies. (4) Investigation of novel transmitter architectures that maximize link loss budget while providing spectrally efficient modulation for data rates of 100Gb\/s\/wavelength.\r\n[\/panel]\r\n\r\n[panel header=\"University College London | Optics for Distributed Learning: Challenges and Opportunities\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UCL PI:<\/strong> <a href=\"https:\/\/www.ucl.ac.uk\/electronic-electrical-engineering\/people\/dr-georgios-zervas\" target=\"_blank\" rel=\"noopener\">Georgios Zervas<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/pcosta\/\">Paolo Costa<\/a><\/p>\r\n<strong>UCL Student:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/alessandro-ottino-09ba1a14b\/\" target=\"_blank\" rel=\"noopener\">Alessandro Ottino<\/a>\r\n\r\nIn the last few years the demand for Machine Learning (ML) and Artificial Intelligence (AI) has experienced an unprecedented growth due to the improvement and availability of hardware accelerators. To increase the prediction capabilities and applications of deep learning models, deeper and larger models trained with larger amount of data are needed. The requirements of these new models is outpacing the hardware improvement, forcing scientists to rely on distributing the ML workload onto multiple accelerators instead of a single one. Scaling these distributed-training workloads is often limited by communication\/ computation time imbalance, making the network one of the key bottlenecks. The goal of this project is to investigate how novel optics technologies could address these challenges by rethinking the network from the grounds up and purposely optimizing it for distributed training.\r\n[\/panel]\r\n\r\n[panel header=\"University College London | Synchronous Subnanosecond Clock and Data Recovery for Optically Switched Data Centres Using Clock Phase Caching\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UCL PI:<\/strong> <a href=\"https:\/\/www.zhixinliu.net\/\" target=\"_blank\" rel=\"noopener\">Zhixin Liu<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PIs:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/hiballan\/\">Hitesh Ballani<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/pcosta\/\">Paolo Costa<\/a><\/p>\r\n<strong>UCL Student:<\/strong> <a href=\"https:\/\/www.ucl.ac.uk\/iccs\/kari-clark\" target=\"_blank\" rel=\"noopener\">Kari Clark<\/a>\r\n\r\nThe rapid growth in the amount of data being transferred within data centres, combined with the slowdown in Moore\u2019s Law, creates challenges for the future scalability of electronically switched data-centre networks. Optical switches could offer a future-proof alternative, and photonic integration platforms have been demonstrated with nanosecond-scale optical switching times. End-to-end switching time is, however, currently limited by the clock and data recovery time, which typically takes microseconds, removing the benefits of nanosecond optical switching. Here we show that a clock phase caching technique can provide clock and data recovery times of under 625\u2009ps (16 symbols at 25.6\u2009Gb\u2009s\u22121). Our approach uses the measurement and storage of clock phase values in a synchronized network to simplify clock and data recovery versus conventional asynchronous approaches. We demonstrate the capabilities of our technique using a real-time prototype with commercial transceivers and validate its resilience against temperature variation and clock jitter.\r\n[\/panel]\r\n[panel header=\"University of California, Santa Barbara | Frequency Stabilized Lasers for Coherent Optical (FRESCO) WDM DSP-Free Datacenter Interconnects\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UCSB PI:<\/strong> <a href=\"https:\/\/iee.ucsb.edu\/people\/daniel-blumenthal\" target=\"_blank\" rel=\"noopener\">Dan Blumenthal<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/hiballan\/\">Hitesh Ballani<\/a><\/p>\r\n<strong>UCSB Students:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/grant-brodnik-42b807a1\/\" target=\"_blank\" rel=\"noopener\">Grant Brodnik<\/a>, <a href=\"https:\/\/ocpi.ece.ucsb.edu\/people-2\/\" target=\"_blank\" rel=\"noopener\">Mark Harrington<\/a>\r\n\r\nAs data center ethernet switches scale toward 100 Tbps, the fiber interconnect will face significant cost, power, and engineering barriers. We describe FRESCO, a fiber data center interconnect (DCI) that brings high order coherent WDM Terabit links inside the data center without the need for DSP and other power consuming technologies. FRESCO replaces high power DSP- and PLL-based carrier phase recovery by employing integrated ultra-low linewidth stimulated Brillouin scattering (SBS) 1550nm lasers stabilized to reference \u00b5-cavities for ~1Hz fundamental linewidth and 10s of Hz integral linewidth optical carriers. This enables the optical frequency stabilized phase locked loop (OFS-PLL) using low bandwidth (&lt;1MHz), low power simple electronics to achieve a direct optical carrier lock with residual phase error variance of 3x10-4 rad2, sufficient for high-order QAM. The architecture is extended to shared source WDM operation using Kerr soliton optical frequency comb generators, amortizing cost and complexity over all channels. Fully integrated silicon photonic transceivers offer reduced footprint and power consumption.\r\n\r\nTo date, we have demonstrated a discrete component single channel 50Gbd 16-QAM FRESCO link using the DSP-free OFS-PLL with low bandwidth (&lt;1MHz), low power electronics for carrier phase recovery over a 200m fiber channel achieving EVM performance within ~1% of a DSP based reference link. We have also demonstrated generation of an ultra-low linewidth WDM shared source using the frequency stabilized SBS laser to pump a Kerr soliton in a SiN microring. Operation as a shared source architecture offers unique opportunities for power savings and motivates current efforts in modeling the impact of stable, ultra-low phase noise optical sources on link performance and power consumption. With our proposed DSP-free architecture, demands on data-chain electronics such as DACs and ADCs may be relaxed, offering additional power savings for tomorrow\u2019s hyperscale DCIs. Future experimental efforts will scale baudrate, QAM order, and WDM count approaching energy efficiency below 10pJ \/bit and 10s of Terabits link capacity.\r\n[\/panel]\r\n[panel header=\"University of Cambridge | Optical Interconnection Architectures for Multi-Chip Systems\"]\r\n<p style=\"margin-bottom: 0px\"><strong>CU PIs:<\/strong> <a href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/rvp11\" target=\"_blank\" rel=\"noopener\">Richard Penty<\/a>, <a href=\"http:\/\/www-g.eng.cam.ac.uk\/photonic_comms\/files\/cps_groupmembers_page.php?crsid=nb301\" target=\"_blank\" rel=\"noopener\">Nikos Bamiedakis<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fokarino\/\">Fotini Karinou<\/a><\/p>\r\n<strong>CU Student:<\/strong> <a href=\"https:\/\/www.linkedin.com\/in\/arastu-sharma-13564ab2\/\" target=\"_blank\" rel=\"noopener\">Arastu Sharma<\/a>\r\n\r\nIncorporating optics in close proximity to high-performance electronic chips on the board-level can not only allow substitution of the existing electrical connections with faster, more power-efficient alternatives, but most importantly, the radical re-design of the system architecture and yet has had limited exploration. Performance of optical links have interdependence on the link utilisation by the application traffic, which makes optical design space a multi-dimensional challenge. We need a solution that looks at the problem from different perspectives, and considers all aspects including the allocation of system resources, technologies and network topology for different applications. Through our study we want to develop a tool that can help in quantifying these design trade-offs, which will in turn help us in designing optimal architectures functioning with high bandwidth, energy efficient optical links. In this project therefore, we study various board-level interconnection architectures for multiple computing chips using on-board optics taking into account recent achievements in the field. To conduct a performance metric simulation, we input our designs into an architecture simulator in combination of an opto-electronic device modelling tool and run for standard workload benchmarks. We compare our design trade-offs with respect to each architecture based on the different available technologies and derive the optimum configuration for execution time, power consumption and implementation complexity.\r\n[\/panel]\r\n[panel header=\"University of Cambridge | Phase and Polarization Spatial Light Modulator (PPSLM) System for Optical Data Storage in Glass\"]\r\n<p style=\"margin-bottom: 0px\"><strong>CU PI:<\/strong> <a href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/dpc31\" target=\"_blank\" rel=\"noopener\">Daping Chu<\/a><\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI: <\/strong><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jaclegg\/\">James Clegg<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/angeor\/\">Andreas Georgiou<\/a><\/p>\r\n<strong>CU Student:<\/strong> <a href=\"http:\/\/www.eng.cam.ac.uk\/profiles\/jh2101\" target=\"_blank\" rel=\"noopener\">Jintao Hong<\/a>\r\n\r\nA low-cost Liquid Crystal display can modulate the polarization of more than 2 billion of pixels per second. These devices are everywhere, from our smartphones, digital cameras and smart watches and provide the highest electro-optical modulation bandwidth in consumer products. When used in a birefringence-based optical data storage system (like Project Silica), they can enable highly parallel data writing making opto-mechanics simpler and relaxing the femtosecond laser requirements. Nevertheless, there are still many challenges including (a) splitting the laser beam and controlling the polarization of each beam (b) controlling the amplitude of the un-modulated light and (c) ensuring the multiple beams are all tightly focused in the glass.\r\n[\/panel]\r\n[panel header=\"University of Oxford | Phase Change Photonic Computing \u2013 from primitives to systems\"]\r\n<p style=\"margin-bottom: 0px\"><strong>UO PI:<\/strong> Harish Bhaskaran<\/p>\r\n<p style=\"margin-bottom: 0px\"><strong>Microsoft PI:<\/strong> Francesca Parmigiani<\/p>\r\nThe use of functional materials that can accumulate information to carry out both memory and computing tasks in-situ is a growing field. Using optical integration onto silicon chips provides a unique opportunity to combine the benefits of silicon scaling and the wavelength multiplexing of optics onto a single platform. In this proposal, we will significantly expand our initial work on carrying out non-von Neumann computations (such as Vector-Matrix Multiplication directly in hardware) to a larger matrix to prove a lab scale demonstration of the potential for such hardware in real-world computing tasks. In addition, this proposal will aim to benchmark such computations against existing state-of-the-art in terms of energy and operational speed.\r\n[\/panel]\r\n[\/accordion]"}],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/679395"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-group"}],"version-history":[{"count":39,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/679395\/revisions"}],"predecessor-version":[{"id":1012338,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/679395\/revisions\/1012338"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/682338"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=679395"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=679395"},{"taxonomy":"msr-group-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group-type?post=679395"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=679395"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=679395"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}