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:20230124T171522Z
LOCATION:C141
DTSTART;TZID=America/Chicago:20221114T110000
DTEND;TZID=America/Chicago:20221114T112500
UID:submissions.supercomputing.org_SC22_sess437_ws_scc107@linklings.com
SUMMARY:Assessing the Current State of AWS Spot Market Forecastability
DESCRIPTION:Workshop\n\nAssessing the Current State of AWS Spot Market For
 ecastability\n\nCaton, Baughman, Haas, Chard, Foster...\n\nSince 2009, Ama
 zon has offered its unused compute capacity as AWS Spot Instances. For the
  first eight years of spot, pure market dynamics and high pricing variabil
 ity created an ideal environment for time-series prediction. Following a p
 ricing-scheme change in 2017, this extreme variability was removed as pric
 ing is artificially smoothed for the end-user, therefore making it signifi
 cantly easier to accurately predict prices. Nevertheless, the literature d
 emonstrates ongoing efforts to accurately predict spot prices. To show pre
 diction in the modern spot market is unnecessary, we train nearly 2.2 mill
 ion ARIMA models on new and old data to demonstrate an order of magnitude 
 improvement in accuracy for models trained on new data. Further, we show t
 his new ease of price prediction makes spot instances ideal for large-scal
 e, cost-aware cloud computing, as cost estimation is now trivial. Accordin
 gly, we demonstrate that even naive prediction approaches waste less than 
 $360 for 1,000,000 core hours.\n\nSession Format: Recorded\n\nRegistration
  Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR