{"id":442950,"date":"2024-07-17T20:28:20","date_gmt":"2024-07-18T03:28:20","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-group&#038;p=442950"},"modified":"2024-09-09T17:33:49","modified_gmt":"2024-09-10T00:33:49","slug":"networking-infrastructure-group","status":"publish","type":"msr-group","link":"https:\/\/www.microsoft.com\/en-us\/research\/group\/networking-infrastructure-group\/","title":{"rendered":"Networking Infrastructure Group"},"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-grey card-background--full-bleed\">\n\t\t\t\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 \">\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\t\t\t\t\t\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/lab\/microsoft-research-asia\/\" class=\"icon-link icon-link--reverse mb-2\" data-bi-cN=\"Return to Microsoft Research Lab - Asia\">\n\t\t\t\t\t\t\t\t\t<span class=\"c-glyph glyph-chevron-left\" aria-hidden=\"true\"><\/span>\n\t\t\t\t\t\t\t\t\tReturn to Microsoft Research Lab &#8211; Asia\t\t\t\t\t\t\t\t<\/a>\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=\"networking-infrastructure-group\">Networking Infrastructure Group<\/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>The Networking Infrastructure Group (NIG) at Microsoft Research Asia (MSRA) engages in fundamental research on all aspects of computer networking and infrastructure. Our group members&#8217; research interests span cloud computing, datacenter networks, wireless, and mobile systems. We investigate new paradigms to build intelligent cloud infrastructure and networks. We emphasize both analytical and empirical approaches, and build running systems to validate theoretical results and refine system principles from experimental work.<\/p>\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 aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"300\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-07-300x300-1.png\" alt=\"cloud outline icon\" class=\"wp-image-878205\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-07-300x300-1.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-07-300x300-1-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-07-300x300-1-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/figure>\n\n\n\n<p class=\"has-text-align-center\"><strong>Next-Gen Data Center Infrastructure<\/strong><\/p>\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 aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"300\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-09-300x300-1.png\" alt=\"microchip outline icon\" class=\"wp-image-878190\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-09-300x300-1.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-09-300x300-1-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-09-300x300-1-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/figure>\n\n\n\n<p class=\"has-text-align-center\"><strong>Next-Gen AI <font face=\"wf_segoe-ui_bold, wf_segoe-ui_semibold, wf_segoe-ui_normal, Arial, sans-serif\">Infrastructur<\/font>e<\/strong><\/p>\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 aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"300\" height=\"300\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-03-300x300-1.png\" alt=\"multi-connected dots icon\" class=\"wp-image-878193\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-03-300x300-1.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-03-300x300-1-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/09\/element-03-300x300-1-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/figure>\n\n\n\n<p class=\"has-text-align-center\"><strong>AI for System and Networking<\/strong><\/p>\n<\/div>\n<\/div>\n\n\n\n<p><\/p>\n\n\n\n<p>Get more information by checking our <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/group\/networking_research_group\/current-projects\/\">current projects<\/a>.<\/p>\n\n\n\n\n\n<p><strong>Next-Gen Data Center Infrastructure<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Hardware-assisted Virtualization Disaggregation\n<ul class=\"wp-block-list\">\n<li>Terminus: The key idea of Terminus is to organize all kinds of resources to form a unified resource pool in the cloud, and all the resources are partitioned and virtualized. By this means, cloud providers could improve the resource utilization for its public cloud, and our customers will experience bare metal performance with the highest security level.<\/li>\n\n\n\n<li>DUA: DUA is a unified framework for FPGA applications to access all resources in data center without CPU involvement by leveraging existing FPGA and networking infrastructure. With DUA, FPGA applications are provided a unified address format and a single set of communication API to access all resources, regardless of the location (remote or local), or the type of the target device (CPU, GPU, FPGA DRAM, server DRAM, SSD, etc.).<\/li>\n\n\n\n<li>S-Direct: It is a solution for performant and transparent flash storage disaggregation. The system is based on FPGA-based SmartNIC, with no dependency on remote resources other than the storage itself and the network bandwidth. It introduces a novel approach to offload the NVMe data path to SmartNICs, providing remote access latency almost identical to local. It also supports an efficient and precise QoS scheduling technique that minimizes performance impact on shared flash. Currently, S-Direct is deployed in real-world applications.<\/li>\n\n\n\n<li>SDR: Software-defined rack (SDR) is a research sandbox for rack-scale focused disaggregation. We designed and assembled SDR hardware, which uses a commodity PCIe\/CXL switch network to allow resource disaggregation at rack-scale and ensure microsecond or sub-microsecond access latencies. In addition, we aim to design and develop a system solution, called RackOS, for the SDR hardware, to address many disaggregation challenges, such as achieve runtime resource allocation\/deallocation, transparent fail-over, and fine-grained QoS.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>Programmable Networking\n<ul class=\"wp-block-list\">\n<li>SmartToR: This project aims at make ToR switches smart and work together to offload cloud applications.<\/li>\n\n\n\n<li>SmartNIC Service: SmartNICs are broadly deployed in today\u2019s data centers due to their performance, energy efficiency, and programmabil- ity benefits. They support a wide variety of offloads to serve host applications, from network functions to distributed ap- plications. However, the tight coupling between SmartNICs and host machines makes it difficult to scale SmartNIC re- sources and causes resource underutilization. To address these issues, we articulate a vision of providing SmartNIC as a service.<\/li>\n\n\n\n<li>SONiC Chassis<\/li>\n\n\n\n<li>SONiC Web App: We are designing a new application type for SONiC called SWA (SONiC Web Application). It has Javascript-like grammar with built-in SONiC libraries support and covers both the control plane and data plane. It contains a customized compiler, a run-time daemon inside SONiC OS and an IDE based on VSCode. It will dramatically reduce the learning curve for SONiC development and boost the SONiC&#8217;s popularity in the NOS (Network Operating System) area.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>Intelligent Host Networking\n<ul class=\"wp-block-list\">\n<li>NetKernel: It decouples the network stack from the guest virtual machine and offers it as an independent module. NetKernel represents a new paradigm where network stack can be managed as part of the virtualized infrastructure. It provides important efficiency benefits: By gaining control and visibility of the network stack, operator can perform network management more directly and flexibly, such as multiplexing VMs running different applications to the same network stack module to save CPU. Users also benefit from the simplified stack deployment and better performance. <\/li>\n\n\n\n<li>PipeDevice: It is a new hardware-software co-design approach for low overhead intra-host container communication.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<p><strong>Next-Gen AI Infrastructure<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Performance\n<ul class=\"wp-block-list\">\n<li><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/tutel\" target=\"_blank\" rel=\"noreferrer noopener\">Tutel<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>: Tutel is a high-performance Mixture-of-Experts (MoE) library to facilitate the development of large-scale DNN models and aim to be the future foundation for extreme-large model training and inference. Tutel is the first MoE framework to enable 4,096 A100 MoE training on Azure. With diverse and flexible MoE algorithmic supported by Tutel, developers across AI domains can execute MoE easier and more efficiently. Tutel has been open sourced at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/tutel\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/github.com\/microsoft\/tutel<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> and part of whose features have been integrated into PyTorch Fairseq, ORT MoE, DeepSpeed, etc.<\/li>\n\n\n\n<li>MSCCL: It is an inter-accelerator communication framework that is built on top of NCCL and uses its building blocks to execute custom-written collective communication algorithms. MSCCL vision is to provide a unified, efficient, and scalable framework for executing collective communication algorithms across multiple accelerators.<\/li>\n\n\n\n<li>ARK: It presents a software-hardware co-design to implement an AI framework that is autonomously executed by GPUs without CPU involvement.<\/li>\n\n\n\n<li>MAGE: It aims to providing a deep insight on memory-side bottlenecks of AI workloads and proposing hardware-agnostic metrics for evaluating performance of memory subsystems in AI accelerators.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>Reliability\n<ul class=\"wp-block-list\">\n<li>SuperBench: It is an automation system for performance validation and diagnosis in Azure AI infrastructure. It is used to guarantee the hardware quality before delivering VMs to Azure customers and identify defects with hardware failures or performance regression, which has helped Azure in shipping qualified emerging GPU SKUs including A100, MI200, etc. SuperBench has been open sourced at&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/superbench\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/aka.ms\/superbench<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, aiming to provide a performance contract for AI hardware among customers, cloud providers, and hardware vendors.<\/li>\n\n\n\n<li>Moneo: Moneo is a non-intrusive cloud-friendly monitoring system, which is capable of intelligently collecting the key architecture-level metrics at finer granularity in real-time without instrumenting or tracing the workloads. Moneo has been open sourced at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/moneo\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/aka.ms\/moneo<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> for Azure customers.<\/li>\n\n\n\n<li>NPKit: It is an online automatic detection and diagnosis framework for communication collective library (NVIDIA NCCL, AMD RCCL and <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/msccl\" target=\"_blank\" rel=\"noreferrer noopener\">Microsoft MSCCL<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, etc.) used in large-scale AI workload. It aims to provide non-intrusive lightweight solution to quickly find the issues on node or\/and IB including NIC regression, port flapping, latency inflation, etc. With NPKit, users can quickly figure out the root cause and resume their large-scale job. NPKit outputs detailed timeline in Google Trace Event Format, by which users can leverage trace viewer to deeply understand and analyze their workflow behavior.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>Prediction\n<ul class=\"wp-block-list\">\n<li>AISim: It can provide additional support throughout the life cycle of our AI products and services, especially from design to operation, to ensure that they comply with the fast-increasing carbon neutral requirements while maintaining huge business impact growth.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<p><strong>AI for System and Networking<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>AI for Security\n<ul class=\"wp-block-list\">\n<li>Bot Detection: One of MSRA-NRG&#8217;s goals is to improve the precision and recall of web crawler (bot) detection so that our customer&#8217;s website data will not be compromised. We have been working on various solutions to detect the bots, including traditional rule-based bot detection methods based on User-Agent, Client IP and advanced methods based on deep learning models like LSTM, CNN, GCN. Our techniques serve all consumers within Microsoft.&nbsp;<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>AI for Networking\n<ul class=\"wp-block-list\">\n<li>R3Net: <\/li>\n\n\n\n<li>OpenNetLab: It aims to build and provide a distributed networking platform with many collaborative nodes and a common benchmarking dataset (i.e. ImageNet in the networking area) for researchers to collect real networking data and train\/evaluate their AI models for various networking environments, including the Internet\/cloud, and wireless and mobile networks. Website: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/opennetlab.org\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/opennetlab.org<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n\n\n<p><strong>Data Center Networking &#8212; <em>building efficient and intelligent fabrics for millions of servers<\/em><\/strong><br \/><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-215785 alignleft\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2001\/06\/dcn.jpg\" alt=\"dcn\" width=\"90\" height=\"90\"> Data center networking (DCN) is a crucial part to support efficient operations of online services in the cloud. We investigate paradigm-shifting technologies to build high availability, high reliability and high-performance network fabric for our data centers.<\/p>\n<\/p>\n<p><strong><em>Past Projects:<\/em><\/strong><\/p>\n<ul>\n<li>Cloud-scale Network Function Virtualization<\/li>\n<li>Programming Reconfigurable Hardware for Virtualized Network Functions (<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/project-clicknp\/\">more<\/a>)\n<ul>\n<li>Checkout our&nbsp;ClickNP paper in SIGCOMM 2016!<\/li>\n<\/ul>\n<\/li>\n<li>Multi-path transport protocols for&nbsp;data center networks (<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/fuso-fast-multi-path-transport-loss-recovery-scheme-data-centers-2\/\">more<\/a>)\n<ul>\n<li>Checkout our&nbsp;FUSO paper in ATC 2016!<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<ul>\n<li>Multicast and low-latency routing: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dx.doi.org\/10.1145\/2413176.2413182\">Datacast<\/a> (CoNEXT&#8217;12), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/conferences.sigcomm.org\/co-next\/2013\/program\/p49.pdf\">RDB<\/a>&nbsp;(CoNEXT&#8217;13)<\/li>\n<li>Congestion control:&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/conferences.sigcomm.org\/co-next\/2010\/CoNEXT_papers\/13-Wu.pdf\">ICTCP<\/a> (CoNEXT&#8217;10 &#8211; Best Paper),&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dx.doi.org\/10.1145\/2413176.2413181\">ECN*<\/a> (CoNEXT&#8217;12)<\/li>\n<li>Network management:&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/conferences.sigcomm.org\/sigcomm\/2010\/papers\/sigcomm\/p39.pdf\">Address Configuration<\/a> (SIGCOMM&#8217;10)<\/li>\n<li>Virtualization:&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/conferences.sigcomm.org\/co-next\/2010\/CoNEXT_papers\/15-Guo.pdf\">SecondNet<\/a> (CoNEXT&#8217;10), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.usenix.org\/events\/nsdi11\/tech\/full_papers\/Lu_Guohan.pdf\">ServerSwitch<\/a> (NSDI&#8217;11 &#8211; Awarded Best Paper)<\/li>\n<li>Our pioneer work in DCN, on network architecture: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/doi.acm.org\/10.1145\/1402958.1402968\">DCell<\/a> (SIGCOMM&#8217;08), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/ccr.sigcomm.org\/online\/?q=node\/501\">BCube<\/a> (SIGCOMM&#8217;09)<\/li>\n<\/ul>\n<p><strong>FPGA for the cloud &#8212; <em>empowering the cloud with <b><i>reconfigurable hardware <\/i><\/b><\/em><\/strong><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\" wp-image-268764 alignleft\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2001\/06\/fpga-300x300.jpg\" alt=\"fpga\" width=\"90\" height=\"90\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/fpga-300x300.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/fpga-150x150.jpg 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/fpga-180x180.jpg 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/fpga-360x360.jpg 360w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/fpga.jpg 693w\" sizes=\"(max-width: 90px) 100vw, 90px\" \/>We foresee that the re-configurability and high parallelism of FPGA will fundamentally change the way we are serving online services. We are investigating the cutting-edge technologies for better programming FPGAs and empowering our services in the cloud. We have demonstrated that using FPGA can significantly accelerate Deep Neural Networks in the cloud.<\/p>\n<p><strong><em>Past Projects:<\/em><\/strong><\/p>\n<ul>\n<li>Going Deeper with FPGA (<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/project-fdnn-fpga-based-deep-neural-networks\/\u200e\">more<\/a>)\n<ul>\n<li>Checkout our FPGA &#8217;16 paper on Angel-Eye!<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p><strong>Software radio &#8212; <em>empowering wireless revolution&nbsp;with the magic of software<\/em><\/strong><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignleft wp-image-215791\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2001\/06\/sora-logo.png\" alt=\"sora-logo\" width=\"91\" height=\"91\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/sora-logo.png 152w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/sora-logo-150x150.png 150w\" sizes=\"(max-width: 91px) 100vw, 91px\" \/>We foresee that the increasingly powerful multi-core architecture and parallel data processing will fundamentally change wireless communications. Software-based cognitive wireless systems will become much more powerful, flexible, and efficient compared with existing wireless technologies. The enabling technology is software radio (or software-defined radio, SDR). For the past several years, we have strike to build the best software radio in a very different&nbsp;way, and to develop programming models and tools that will greatly reduce the effort involved in implementing state-of-the-art wireless systems.<\/p>\n<ul>\n<li>Publication: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.usenix.org\/events\/nsdi09\/tech\/full_papers\/tan\/tan.pdf\">Sora<\/a> (NSDI&#8217;09 &#8211; Best Paper, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dx.doi.org\/10.1145\/1866739.1866760\">CACM version<\/a>)<\/li>\n<li>Sora is <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/Microsoft\/Sora\/\">open-source at GitHub<\/a>.<\/li>\n<\/ul>\n<p><strong>Wireless Networks&nbsp;&#8212; <em>computational thinking in&nbsp;the traditional field<\/em><\/strong><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignleft wp-image-215794\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2001\/06\/wireless.png\" alt=\"wireless\" width=\"85\" height=\"93\">The explosive increase in mobile usage&nbsp;demands constant revolutions in wireless technology. We have been engaged in a broad range of wireless research in PHY, MAC, cross-layer, MIMO, dynamic spectrum, whitespace, measurement, reliability, etc. Our approach has been one with&nbsp;&#8220;computational wireless&#8221; &#8212; applying computer science methodology in a very traditional field.<\/p>\n<p><strong><em>Past Projects:<\/em><\/strong><\/p>\n<ul>\n<li>MIMO Systems\n<ul>\n<li>Checkout SmartHub paper at HotNets 2015<\/li>\n<li>BigStation at SIGCOMM 2013<\/li>\n<\/ul>\n<\/li>\n<li>Networking&nbsp;at 60GHz<\/li>\n<li>Software-Defined WLAN: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/ieeexplore.ieee.org\/xpls\/abs_all.jsp?arnumber=7218521&tag=1\">TDMA in WLAN <\/a>(INFOCOM&#8217;15)<\/li>\n<li>MAC and PHY: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dl.acm.org\/citation.cfm?id=2746299\">Pilotfish&nbsp;<\/a>(MobiHoc&#8217;15), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dx.doi.org\/10.1145\/2348543.2348557\">uACKs<\/a> (MobiCom&#8217;12), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/conferences.sigcomm.org\/sigcomm\/2010\/papers\/sigcomm\/p147.pdf\">Fine-grained Channel Access<\/a><b> <\/b>(SIGCOMM&#8217;10), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/dx.doi.org\/10.1145\/1859995.1860028\">Retransmission<\/a> (MobiCom&#8217;10), <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/doi.acm.org\/10.1145\/1614320.1614327\">SAM<\/a> (MobiCom&#8217;09)<\/li>\n<li>DSA & Whitespace: <a class=\"invalidLink\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/enable-flexible-spectrum-access-with-spectrum-virtualization\/\">Spectrum Virtualization<\/a> (DySPAN&#8217;12)<\/li>\n<li>Video\/visual: <a href=\"http:\/\/dx.doi.org\/10.1145\/2348543.2348573\">ParCast<\/a> (MobiCom&#8217;12), <a href=\"http:\/\/dx.doi.org\/10.1145\/2500423.2500437\">QuickSync<\/a> (MobiCom&#8217;13)<\/li>\n<\/ul>\n<p><strong>Mobile Systems &#8212; <em>connecting people, devices, and information<\/em> <\/strong><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignleft wp-image-215797\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2001\/06\/mobile-300x267.jpg\" alt=\"mobile\" width=\"101\" height=\"90\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/mobile-300x267.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2001\/06\/mobile.jpg 403w\" sizes=\"(max-width: 101px) 100vw, 101px\" \/>Our&nbsp;mobile systems research covers&nbsp;exploring new paradigms and system mechanisms, and bringing mobile devices and the Cloud together to further improve a user\u2019s mobile experience, and to enable a spectrum of new mobile applications that harness the power of the Cloud.<\/p>\n<\/p>\n<p><strong><em>Past Projects:<\/em><\/strong><\/p>\n<ul>\n<li>Smart and Secure Home Router.\n<ul>\n<li>Checkout our paper in APSys 2016<\/li>\n<\/ul>\n<\/li>\n<li>Localizations: <a href=\"http:\/\/dl.acm.org\/citation.cfm?id=2742648\">Pixel<\/a> (Mobisys&#8217;15) <a href=\"http:\/\/dx.doi.org\/10.1145\/2426656.2426709\">IODetector<\/a>&nbsp;(SenSys&#8217;12), <a href=\"https:\/\/www.usenix.org\/conference\/nsdi13\/walkie-markie-indoor-pathway-mapping-made-easy\">Walkie-Markie<\/a> (NSDI&#8217;13), <a href=\"https:\/\/www.usenix.org\/conference\/nsdi14\/technical-sessions\/presentation\/li\">Epsilon<\/a>&nbsp;(NSDI&#8217;14).<\/li>\n<li>Power efficiency: <a href=\"http:\/\/dx.doi.org\/10.1145\/2307636.2307675\">DozyAP<\/a>&nbsp;&#8211; tethering&nbsp;(MobiSys&#8217;12), <a href=\"https:\/\/www.usenix.org\/conference\/nsdi13\/v-edge-fast-self-constructive-power-modeling-smartphones-based-battery-voltage\">V-edge<\/a> &#8211;&nbsp;power modelling (NSDI&#8217;13), <a href=\"http:\/\/dx.doi.org\/10.1145\/2462456.2464444\">Email Sync<\/a> (MobiSys&#8217;13)<\/li>\n<li>mobile&nbsp;hardware: <a href=\"http:\/\/dx.doi.org\/10.1145\/2426656.2426662\">MusicalHeart<\/a> (SenSys&#8217;12),&nbsp;<a href=\"http:\/\/dx.doi.org\/10.1145\/2462456.2464446\">Auditeur<\/a> (MobiSys&#8217;13), <a href=\"http:\/\/dx.doi.org\/10.1145\/2462456.2464454\">ViRi<\/a>&nbsp;(MobiSys&#8217;13)<\/li>\n<li>Sensing people: <a href=\"http:\/\/dx.doi.org\/10.1145\/2462456.2464449\">Mood Sensor<\/a> (MobiSys&#8217;13)<\/li>\n<li>Mobile collaborations: <a href=\"http:\/\/doi.acm.org\/10.1145\/1247660.1247667\">Mobius<\/a> (MobiSys&#8217;07), <a href=\"http:\/\/doi.acm.org\/10.1145\/1322263.1322265\">BeepBeep<\/a> (SenSys&#8217;07), <a href=\"http:\/\/doi.acm.org\/10.1145\/1555816.1555831\">Point&Connect<\/a> (MobiSys&#8217;09)<\/li>\n<li>OS and chips: <a href=\"http:\/\/doi.acm.org\/10.1145\/1555816.1555819\">xShare<\/a> (MobiSys&#8217;09),&nbsp;<a href=\"http:\/\/doi.acm.org\/10.1145\/1614320.1614345\">Micro Wireless Interconnect<\/a> (MobiCom&#8217;09)<\/li>\n<\/ul>\n\n\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Best Paper Award<\/strong>, APNet &#8217;22 (OpenNetLab)<\/li>\n\n\n\n<li><strong>Best Paper Award<\/strong>, MLArchSys&#8217;22 (GPU-driven Code Execution)<\/li>\n\n\n\n<li><strong>Best Paper Award<\/strong>, NSDI&#8217;11 (ServerSwitch)<\/li>\n\n\n\n<li><strong>Best Paper Award<\/strong>, CoNEXT&#8217;10 (Incast)<\/li>\n\n\n\n<li>Best Demo Award, SIGCOMM&#8217;10 (Sora\/MPAP)<\/li>\n\n\n\n<li><strong>Best Paper Award<\/strong>, NSDI&#8217;09 (Sora)<\/li>\n\n\n\n<li>Best Demo Award, NSDI&#8217;09 (Sora)<\/li>\n\n\n\n<li>Best Demo Award, MobiSys&#8217;08 (P&C)<\/li>\n\n\n\n<li>Best Demo Award, SenSys&#8217;07 (BeepBeep)<\/li>\n\n\n\n<li>Best Demo Award, MobiSys&#8217;07 (Mobius)<\/li>\n<\/ul>\n\n\n","protected":false},"excerpt":{"rendered":"<p>The Networking Infrastructure Group at Microsoft Research Asia engages in fundamental research on all aspects of computer networking and infrastructure.<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr_group_start":"2017-11-28","footnotes":""},"research-area":[13547],"msr-group-type":[243694],"msr-locale":[268875],"msr-impact-theme":[],"class_list":["post-442950","msr-group","type-msr-group","status-publish","hentry","msr-research-area-systems-and-networking","msr-group-type-group","msr-locale-en_us"],"msr_group_start":"2017-11-28","msr_detailed_description":"","msr_further_details":"","msr_hero_images":[],"msr_research_lab":[199560],"related-researchers":[{"type":"user_nicename","display_name":"Yongqiang Xiong","user_id":35049,"people_section":"Group Members","alias":"yqx"},{"type":"user_nicename","display_name":"Peng Cheng","user_id":33225,"people_section":"Group Members","alias":"pengc"},{"type":"user_nicename","display_name":"Lei Qu","user_id":32650,"people_section":"Group Members","alias":"lequ"},{"type":"user_nicename","display_name":"Ran Shu","user_id":37926,"people_section":"Group Members","alias":"rashu"},{"type":"user_nicename","display_name":"Zhixiong Niu","user_id":38118,"people_section":"Group Members","alias":"zhniu"},{"type":"user_nicename","display_name":"Ziyue Yang","user_id":41653,"people_section":"Group Members","alias":"ziyyang"},{"type":"user_nicename","display_name":"Yifan Xiong","user_id":40702,"people_section":"Group Members","alias":"yixio"},{"type":"user_nicename","display_name":"Changho Hwang","user_id":41844,"people_section":"Group Members","alias":"changhohwang"},{"type":"user_nicename","display_name":"Rui Gao","user_id":33465,"people_section":"Group Members","alias":"ruigao"},{"type":"user_nicename","display_name":"Yuting Jiang","user_id":41868,"people_section":"Group Members","alias":"yutingjiang"},{"type":"user_nicename","display_name":"Tao Zhang","user_id":42267,"people_section":"Group Members","alias":"zhangt"},{"type":"user_nicename","display_name":"Wenxue Cheng","user_id":43500,"people_section":"Group Members","alias":"wenxchen"},{"type":"user_nicename","display_name":"Shuotao Xu","user_id":42015,"people_section":"Collaborators","alias":"shuotaoxu"},{"type":"user_nicename","display_name":"Yuqing Yang","user_id":40654,"people_section":"Collaborators","alias":"yuqyang"}],"related-publications":[544137,507281,494123,431784,507263,506705,309230,442566,860589,873900,897375,1056678],"related-downloads":[860751],"related-videos":[],"related-projects":[],"related-events":[],"related-opportunities":[1059471,1059477,1059615,1059621],"related-posts":[],"tab-content":[],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/442950"}],"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":102,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/442950\/revisions"}],"predecessor-version":[{"id":1083729,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group\/442950\/revisions\/1083729"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=442950"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=442950"},{"taxonomy":"msr-group-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-group-type?post=442950"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=442950"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=442950"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}