BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201806Z
LOCATION:242
DTSTART;TZID=America/Chicago:20251116T104500
DTEND;TZID=America/Chicago:20251116T111000
UID:submissions.supercomputing.org_SC25_sess216_ws_isav110@linklings.com
SUMMARY:From Exploration to Explanation: ML-Driven Causal Discovery for Da
 tacenter Reliability at Scale
DESCRIPTION:Pavana Prakash, Rolando P. Hong Enriquez, and Sergey Serebryak
 ov (Hewlett Packard Labs); David Grant and Wesley Brewer (Oak Ridge Nation
 al Laboratory (ORNL)); and Dejan Milojicic (Hewlett Packard Labs)\n\nModer
 n datacenters operate at unprecedented scale, supporting HPC and AI worklo
 ads while consuming hundreds of megawatts of power. Their reliability is c
 hallenged by complex interdependencies across cooling, power, and network 
 subsystems, where failures can cascade into downtime and degraded performa
 nce. Existing monitoring approaches, largely threshold or only correlation
 -based, struggle to isolate root causes within high-dimensional, evolving 
 telemetry. We present PACE (Pattern and Causal Exploration), an ML-based f
 ramework that combines unsupervised correlation clustering with supervised
 , lag-aware Granger causality to uncover subsystem structure and directed 
 causal pathways from multivariate telemetry. PACE yields interpretable cau
 sal graphs, subsystem heatmaps, that align with physical processes and con
 trol logic, providing actionable insights for operations. Finally, we disc
 uss how embedding PACE into digital twin architectures enables causal-info
 rmed \emph{what-if} reasoning, advancing the reliability and efficiency in
  datacenters.\n\nTag: AI, Machine Learning, & Deep Learning, Clouds & Dist
 ributed Computing, Performance Evaluation, Scalability, & Portability, Sci
 entific & Information Visualization\n\nRecording: Partially Livestreamed, 
 Partially Recorded\n\nRegistration Category: Technical Program Reg Pass, W
 orkshop Reg Pass\n\nSession Chairs: Axel Huebl (Lawrence Berkeley National
  Laboratory (LBNL)); Will Usher (Luminary Cloud Inc); Nicola Ferrier (Argo
 nne National Laboratory (ANL), University of Chicago); Estelle Dirand (Tot
 alEnergies, France); Victor Mateevitsi (Argonne National Laboratory (ANL),
  University of Illinois Chicago); and Jens Henrik Göbbert (Forschungszentr
 um Jülich, Jülich Supercomputing Centre (JSC))\n\n
END:VEVENT
END:VCALENDAR