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:20230124T171520Z
LOCATION:D172
DTSTART;TZID=America/Chicago:20221118T091000
DTEND;TZID=America/Chicago:20221118T093500
UID:submissions.supercomputing.org_SC22_sess449_ws_resdis101@linklings.com
SUMMARY:Methodology for Evaluating the Potential of Disaggregated Memory S
 ystems
DESCRIPTION:Workshop\n\nMethodology for Evaluating the Potential of Disagg
 regated Memory Systems\n\nDing, Williams, Nam, Groves, Awan...\n\nTightly-
 coupled HPC systems have rigid memory allocation and can result in expensi
 ve memory resource under-utilization. As novel memory and network technolo
 gies mature, disaggregated memory systems are becoming a promising solutio
 n for future HPC systems. It allows workloads to use the available memory 
 of the entire system. We propose a design framework to explore the disaggr
 egated memory system design space. The framework incorporates memory capac
 ity, network bandwidth, and local and remote memory access ratio, and prov
 ides an intuitive approach to guide machine configurations based on techno
 logy trends and workload characteristics. We apply our framework to analyz
 e eleven workloads from five computational scenarios, including AI trainin
 g, data analysis, genomics, protein, and traditional HPC. We demonstrate t
 he ability of our methodology to understand the potential and pitfalls of 
 a disaggregated memory system and motivate machine configurations. Our met
 hodology shows that 10 out of our 11 applications/workflows can leverage d
 isaggregated memory without affecting performance.\n\nSession Format: Reco
 rded\n\nTag: AI-HPC Convergence, Emerging Technologies, Memory Systems, Ne
 tworks, Resource Management and Scheduling\n\nRegistration Category: Works
 hop Reg Pass
END:VEVENT
END:VCALENDAR