BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20230312T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20231105T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20231121T223141Z UID:05BB6230-7B38-4D74-8A72-3C54390924EB DTSTART;TZID=America/New_York:20231101T094500 DTEND;TZID=America/New_York:20231101T103000 DESCRIPTION:Network management is essential for maintaining high performanc e\, security\, and reliability in large scale networks. Network management involves measuring network state\, making decisions for various managemen t applications\, and enforcing the decisions into individual devices. In l ieu of recent progress in machine learning\, one natural question is how s hould we build network management systems to better leverage these emergin g techinques.\n\nIn this talk\, I will discuss three key challenges in bui lding future management systems driven by ML: How to collect a large amoun t of diverse data to feed into ML systems\, how to enable fast and accurat e management decisions using ML\, and how to enable fast reaction to netwo rk events. I'll show a few example systems we built to address these chall enges: The first system is DTA\, a direct telemetry access framework that aggregates and moves hundreds of millions of reports per second from switc hes into queryable data structures in collectors’ memory\; the second sy stem is Teal\, a learning-based algorithm that leverages the parallel proc essing power of GPUs to accelerate traffic engineering control\; finally\, I'll discuss our preliminary ideas on enabling fast reactions to network events.\n\nSpeaker(s): Minlan Yu\n\nNiagara Falls\, Ontario\, Canada LOCATION:Niagara Falls\, Ontario\, Canada ORGANIZER:rboutaba@uwaterloo.ca SEQUENCE:14 SUMMARY:Network Management in the Machine Learning Era URL;VALUE=URI:https://events.vtools.ieee.org/m/385223 X-ALT-DESC:Description: <br /><p>Network management is essential for mainta ining high performance\, security\, and reliability in large scale network s. Network management involves measuring network state\, making decisions for various management applications\, and enforcing the decisions into ind ividual devices. In lieu of recent progress in machine learning\, one natu ral question is how should we build network management systems to better l everage these emerging techinques.</p>\n<p>In this talk\, I will discuss t hree key challenges in building future management systems driven by ML: Ho w to collect a large amount of diverse data to feed into ML systems\, how to enable fast and accurate management decisions using ML\, and how to ena ble fast reaction to network events. I'll show a few example systems we bu ilt to address these challenges: The first system is DTA\, a direct teleme try access framework that aggregates and moves hundreds of millions of rep orts per second from switches into queryable data structures in collectors &rsquo\; memory\; the second system is Teal\, a learning-based algorithm t hat leverages the parallel processing power of GPUs to accelerate traffic engineering control\; finally\, I'll discuss our preliminary ideas on enab ling fast reactions to network events.</p> END:VEVENT END:VCALENDAR