BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210527T134950Z
LOCATION:Applaus
DTSTART;TZID=Europe/Stockholm:20200622T220600
DTEND;TZID=Europe/Stockholm:20200622T224400
UID:isc_hpc_ISC High Performance 2020_sess328_phd101@linklings.com
SUMMARY:Leveraging LSTMs for interference-aware run-time system Predictabi
 lity of cloud workloads
DESCRIPTION:PhD Forum\n\nLeveraging LSTMs for interference-aware run-time 
 system Predictability of cloud workloads\n\nMasouros, Xydis, Soudris\n\nMo
 dern micro-service and container-based cloud-native applications have leve
 raged multi-tenancy as a first class system design concern. The increasing
  number of co-located services/workloads into server facilities stresses r
 esource availability and system capability in an unconventional and unpred
 ictable manner.\nTo efficiently manage resources in such dynamic environme
 nts, run-time observability and forecasting are required to capture worklo
 ad sensitivities under differing interference effects, according to applie
 d co-location scenarios.\nWhile several research efforts have emerged on i
 nterference-aware performance modelling, they are usually applied at a ver
 y coarse-grained manner e.g. estimating the overall performance degradatio
 n of an application, thus failing to effectively quantify, predict or prov
 ide educated insights on the impact of continuous runtime interference on 
 per-resource allocations.\nIn this work, we present a predictive monitorin
 g system that leverages the power of Long Short-Term Memory networks to en
 able fast and accurate runtime forecasting of key performance metrics and 
 resource stresses of cloud-native applications under interference.\nWe eva
 luate our approach under a diverse set of interference scenarios for a ple
 thora of representative cloud workloads, showing that i) we achieve extrem
 ely high prediction accuracy, average R^2 value of 0.98, ii) enable very d
 eep prediction horizons retaining high accuracy, e.g. R^2 of around 0.99 f
 or a horizon of 1 sec ahead and around 0.94 for an horizon of 5 sec ahead,
  while iii) satisfying, at the same time, the strict latency constraints r
 equired to make our proposed framework practical for continuous predictive
  monitoring at runtime.\n\nTag: Pre-Recorded
URL:https://2020.isc-program.com/presentation/?id=phd101&sess=sess328
END:VEVENT
END:VCALENDAR

