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:20211207T055404Z
LOCATION:223
DTSTART;TZID=America/Chicago:20211115T090000
DTEND;TZID=America/Chicago:20211115T165500
UID:submissions.supercomputing.org_SC21_sess332@linklings.com
SUMMARY:PDSW: Sixth International Parallel Data Systems Workshop
DESCRIPTION:Workshop\n\nPDSW Lunch Break (12:30-2)\n\n\n\n----------------
 -----\nPDSW:  Morning Break (10-10:30)\n\n\n\n---------------------\nWelco
 me and Introduction\n\nIbrahim, Sato, Chi Zhou\n\n---------------------\nN
 ew Challenges of Benchmarking All-Flash Storage for HPC\n\nLockwood, Chius
 ole, Wright\n\nThe proliferation of extreme-scale analytics and AI have mo
 tivated the creation of new parallel storage systems that either use all f
 lash or combine nonvolatile memory (NVRAM) and flash to achieve the best a
 ttributes of both media.  Clearly such storage systems have relevance in H
 PC, especially to m...\n\n---------------------\nVerifying IO Synchronizat
 ion from MPI Traces\n\nYellapragada, Wang, Snir\n\nThe paper addresses the
  following question: Are IO operations of HPC applications properly synchr
 onized? We focus on parallel file systems that satisfy POSIX semantics. Th
 e outcome of I/O operations is well-defined provided that conflicting acce
 sses to a file location are not concurrent, but are ord...\n\n------------
 ---------\nClosing Remarks\n\n\n\n---------------------\nHigh-Throughput S
 mall File Access for Large-Scale Machine Learning Applications\n\nOhtsuji\
 n\n---------------------\nInvited Industry Talk: Jump's Archive for the Ne
 xt Decade\n\nDavies\n\nHow and why Jump built a high performance (hundreds
  of gigabytes per second) and scalable (hundreds of PB) POSIX interface to
  an object store,​ using the CernVM open source filesystem.\n\n-----
 ----------------\nNetwork-Accelerated Distributed File Systems\n\nDi Girol
 amo\n\n---------------------\nOptimising I/O Using Non-Volatile Memory\n\n
 Jackson\n\n---------------------\nUnderstanding the I/O Impact on the Perf
 ormance of High-Throughput Molecular Docking\n\nMarkidis, Gadioli, Gadioli
 , Palermo\n\nHigh-throughput molecular docking is a data-driven simulation
  methodology to estimate millions of molecules' position and interaction s
 trength (ligands) when interacting with a given protein site. Because of i
 ts data-driven nature, the high-throughput molecular docking performance d
 epends on how fas...\n\n---------------------\nPDSW Afternoon Break (3-3:3
 0)\n\n\n\n---------------------\nSixth International Parallel Data Systems
  Workshop\n\nIbrahim, Sato, Zhou, Lofstead\n\nEfficient data storage and d
 ata management are crucial to scientific productivity in both traditional 
 simulation-oriented HPC environments and Big Data analysis environments. T
 hese issues are further exacerbated by the growing volume of experimental 
 and observational data, the widening gap between t...\n\n-----------------
 ----\nHyperconverged Storage for High Performance Data Analysis in High En
 ergy Physics: A Case of Intel DAOS Deployment\n\nMoskovsky\n\n------------
 ---------\nUser-Centric System Fault Identification Using IO500 Benchmark\
 n\nLiem, Loftstead\n\nI/O performance in a multi-user environment is diffi
 cult to predict. Users do not know what to expect when running and tuning 
 their application for better I/O performance. We propose to use the IO500 
 benchmark as a way to guide user expectations on their application's I/O p
 erformance and identifying...\n\n---------------------\nI/O Bottleneck Det
 ection and Tuning: Connecting the Dots Using Interactive Log Analysis\n\nB
 ez, Tang, Xie, Williams-Young, Latham...\n\nUsing parallel file systems ef
 ficiently is a tricky problem due to inter-dependencies among multiple lay
 ers of I/O software, including high-level I/O libraries (HDF5, netCDF, etc
 .), MPI-IO, POSIX, and file systems (GPFS, Lustre, etc.). Profiling tools 
 such as Darshan collect traces to help understan...\n\n-------------------
 --\nData-Aware Storage Tiering for Deep Learning\n\nXu, Bhattacharya, Folt
 in, Byna, Faraboschi\n\nDNN models trained with large datasets can perform
  rich deep learning tasks with high accuracy. However, feeding huge volume
 s of training data exerts significant pressure on IO subsystems as the ent
 ire data is re-loaded in random order on every iteration to enable converg
 ence, with very little scope...\n\n---------------------\nSCTuner: An Auto
 -Tuner Addressing Dynamic I/O Needs on Supercomputer I/O Sub-Systems\n\nTa
 ng, Xie, Byna, Carns, Koziol...\n\nThis work proposes SCTuner, an auto-tun
 er integrated within I/O library itself to dynamically tune both the I/O l
 ibrary and the underlying I/O stack at application runtime. To this end, w
 e introduce a statistical benchmarking method to profile the behaviors of 
 individual supercomputer I/O sub-system...\n\n---------------------\nInvit
 ed Talk:  Why Memory Is Your Next Bottleneck and How To Overcome It\n\nAgu
 ilera\n\n---------------------\npMEMCPY: A Simple, Lightweight, and Portab
 le I/O Library for Storing Data in Persistent Memory\n\nLogan\n\n\nTag: Da
 ta Analytics, Data Management, File Systems and I/O, Storage\n\nRegistrati
 on Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR