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:20181221T160726Z
LOCATION:D170
DTSTART;TZID=America/Chicago:20181111T153000
DTEND;TZID=America/Chicago:20181111T155200
UID:submissions.supercomputing.org_SC18_sess149_ws_mchpc112@linklings.com
SUMMARY:On the Applicability of PEBS-Based Online Memory Access Tracking f
 or Heterogeneous Memory Management at Scale
DESCRIPTION:Workshop\nMemory, NVRAM, Parallel Programming Languages, Libra
 ries, and Models, Workshop Reg Pass\n\nOn the Applicability of PEBS-Based 
 Online Memory Access Tracking for Heterogeneous Memory Management at Scale
 \n\nRoca Nonell, Gerofi, Bautista-Gomez, Martinet, Beltran Querol...\n\nOp
 erating systems have historically had to manage only a single type of memo
 ry device. The imminent availability of heterogeneous memory devices based
  on emerging memory technologies confronts the classic single memory model
  and opens a new spectrum of possibilities for memory management. Transpar
 ent data movement based on access patterns of applications is a desired fe
 ature to hide the complexity of memory management to end users. However, c
 apturing memory access patterns of an application at runtime comes at a co
 st, which is particularly challenging for large scale parallel application
 s that may be sensitive to system noise.In this work, we focus on the acce
 ss pattern profiling. We study the feasibility of using Intel’s Processor 
 Event Based Sampling (PEBS) feature to record memory accesses by sampling 
 at runtime and study the overhead at scale. We have implemented a custom P
 EBS driver in the IHK/McKernel lightweight multi-kernel operating system, 
 one of whose advantages is minimal system interference due to the lightwei
 ght kernel’s simple design compared to other OS kernels such as Linux. We 
 present the PEBS overhead of a set of scientific applications and show the
  access patterns identified in noise sensitive HPC applications. Our resul
 ts show that clear access patterns can be captured with 10% overhead in th
 e worst case when running on up to 128k CPU cores (2,048 Intel Xeon Phi Kn
 ights Landing nodes). We conclude that online memory access profiling usin
 g PEBS at large scale is promising for memory management in heterogeneous 
 memory environments.
URL:https://sc18.supercomputing.org/presentation/?id=ws_mchpc112&sess=sess
 149
END:VEVENT
END:VCALENDAR