BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260610T133909Z UID:47D7B0C7-CAF7-49E3-ABB3-1B0E421DDBA6 DTSTART;TZID=America/New_York:20260616T130000 DTEND;TZID=America/New_York:20260616T140000 DESCRIPTION:The rapid growth of artificial intelligence (AI) workloads is d riving the need for highly scalable\, low-latency\, low-complexity\, resil ient and sustainable data center networks (DCNs). AI training and inferenc e environments is placing significant pressure on IP-based routing archite ctures designed for technologies and communications of decades ago. While substantial efforts focus on expanding AI adoption across domains\, parall el initiatives aimed at mitigating the AI-driven data center energy crisis largely emphasize improved cooling technologies and increased power gener ation capacity. The proposed research project targets the root cause of th ese challenges—the continued reliance on retrofitted\, heavyweight legac y protocols\, which introduce inefficiencies that degrade DCN performance and contribute significantly to their escalating power consumption.\nThe p roject Multi-Root Meshed Tree Protocol (MR-MTP) offers a simple IP agnosti c solution by adopting a novel technique to establish multiple trees roote d at leaf nodes in a folded-Clos topology DCN that virtually mesh at the s pines to provide multiple loop-free forwarding paths between server racks. MR-MTP replaces seven protocols in the traditional IP-based DCN router pr otocol stack\, resulting in heavy reduction in router hardware and softwar e\, leading to highly reduced operational complexity and a proportional re duction in energy consumption and equipment cost. MR-MTP’s auto-configur ation reduces human errors and simplifies network maintenance\, management \, and troubleshooting and the IP agnostic approach mitigates vulnerabilit ies associated with IP-layer attacks.\nAI workloads dictate bypassing trad itional IP/TCP stack\, minimizing software overhead\, designing determinis tic fabrics and collapsing multiple protocols among others. Infiniband and RoCE(Ethernet +Remote Direct Memory Access) are two technologies targetin g novel approaches. While InfiniBand demonstrates the benefits of a non-IP \, simplified\, high-performance fabric\, and RoCE retrofit Ethernet to me et AI demands\, both approaches do not address the root cause—protocol c omplexity. MR-MTP collapses multiple protocols to a single\, topology-awar e mechanism\, enabling simpler\, faster\, and more secure data center netw orks. Hyperscale system challenges directly expose the limitations MR-MTP is trying to eliminate. MR-MTP is the next step beyond RDMA fabrics and IP -based DCNs—where the network is no longer configured through protocols but emerges directly from topology.\nC-coded MR-MTP was tested using FABRI C large-scale experimental testbed to emulate near–real-world deployment scenarios and its performance compared with the popular protocol suite Bo rder Gateway Protocol (BGP)\, Equal Cost Multipath Protocol (ECMP) and Bid irectional Forwarding Detection (BFD) in folded-Clos DCN topologies. Resul ts demonstrate the significant performance improvements and stability of M R-MTP. (Code and FABRIC test scripts available on GitHub).\nMR-MTP support s interoperability with DCNs operating on conventional protocols- allowing incremental deployment. With a Technology Readiness Level (TRL) of 6 and recognition by DARPA’s Expedited Research Implementation Series (ERIS) a s a Breakthrough Technological Advancement\, MR-MTP is positioned for real -world deployment.\n\nSpeaker(s): Nirmala \, \n\nVirtual: https://events.v tools.ieee.org/m/563124 LOCATION:Virtual: https://events.vtools.ieee.org/m/563124 ORGANIZER:nxsvks@rit.edu SEQUENCE:27 SUMMARY:Reinventing Sustainable Data Center Networks for the AI Age URL;VALUE=URI:https://events.vtools.ieee.org/m/563124 X-ALT-DESC:Description: <br /><p>The rapid growth of artificial intelligenc e (AI) workloads is driving the need for highly scalable\, low-latency\, l ow-complexity\, resilient and sustainable data center networks (DCNs). AI training and inference environments is placing significant pressure on IP- based routing architectures designed for technologies and communications o f decades ago. While substantial efforts focus on expanding AI adoption ac ross domains\, parallel initiatives aimed at mitigating the AI-driven data center energy crisis largely emphasize improved cooling technologies and increased power generation capacity. The proposed research project targets the root cause of these challenges&mdash\;the continued reliance on retro fitted\, heavyweight legacy protocols\, which introduce inefficiencies tha t degrade DCN performance and contribute significantly to their escalating power consumption.<br>The project&nbsp\;<strong>Multi-Root Meshed Tree Pr otocol (MR-MTP)</strong>&nbsp\;offers a simple IP agnostic solution by ado pting a novel technique to establish multiple trees rooted at leaf nodes i n a folded-Clos topology DCN that virtually mesh at the spines to provide multiple loop-free forwarding paths between server racks. MR-MTP replaces seven protocols in the traditional IP-based DCN router protocol stack\, re sulting in heavy reduction in router hardware and software\, leading to hi ghly reduced operational complexity and a proportional reduction in energy consumption and equipment cost. MR-MTP&rsquo\;s auto-configuration reduce s human errors and simplifies network maintenance\, management\, and troub leshooting and the IP agnostic approach&nbsp\;mitigates vulnerabilities as sociated with IP-layer attacks.<br>AI workloads dictate bypassing traditio nal IP/TCP stack\, minimizing software overhead\, designing deterministic fabrics and collapsing multiple protocols among others. Infiniband and RoC E(Ethernet +Remote Direct Memory Access) are two technologies targeting no vel approaches. While InfiniBand demonstrates the benefits of a non-IP\, s implified\, high-performance fabric\, and RoCE retrofit Ethernet to meet A I demands\, both approaches do not address the root cause&mdash\;protocol complexity. MR-MTP collapses multiple protocols to a single\, topology-awa re mechanism\, enabling simpler\, faster\, and more secure data center net works. Hyperscale system challenges directly expose the limitations MR-MTP is trying to eliminate. MR-MTP is the next step beyond RDMA fabrics and I P-based DCNs&mdash\;where the network is no longer configured through prot ocols but emerges directly from topology.<br>C-coded MR-MTP was tested usi ng&nbsp\;<strong>FABRIC large-scale experimental testbed</strong>&nbsp\;to emulate near&ndash\;real-world deployment scenarios and its performance c ompared with the popular protocol suite&nbsp\;<strong>Border Gateway Proto col (BGP)\, Equal Cost Multipath Protocol (ECMP) and Bidirectional Forward ing Detection (BFD) i</strong>n folded-Clos DCN topologies. Results demons trate the significant performance improvements and stability of MR-MTP. (C ode and FABRIC test scripts available on GitHub).<br>MR-MTP supports inter operability with DCNs operating on conventional protocols- allowing increm ental deployment. With a&nbsp\;<strong>Technology Readiness Level (TRL) of 6</strong>&nbsp\;and recognition by&nbsp\;<strong>DARPA&rsquo\;s Expedite d Research Implementation Series (ERIS)</strong>&nbsp\;as a&nbsp\;<strong> Breakthrough Technological Advancement</strong>\, MR-MTP is positioned for real-world deployment. &nbsp\;</p> END:VEVENT END:VCALENDAR