in memory computing at scale look beyond physical dram
play

IN-MEMORY COMPUTING AT SCALE? LOOK BEYOND PHYSICAL DRAM! Iacovos G. - PowerPoint PPT Presentation

Jan 24, 2024 •317 likes •445 views

IN-MEMORY COMPUTING AT SCALE? LOOK BEYOND PHYSICAL DRAM! Iacovos G. Kolokasis , Anastasios Papagiannis, Polyvios Pratikakis, and Angelos Bilas October 25, 2019 Institute of Computer Science (ICS) Foundation of Research and T echnology

  1. IN-MEMORY COMPUTING AT SCALE? LOOK BEYOND PHYSICAL DRAM! Iacovos G. Kolokasis , Anastasios Papagiannis, Polyvios Pratikakis, and Angelos Bilas October 25, 2019 Institute of Computer Science (ICS) Foundation of Research and T echnology – Hellas (FORTH) & Computer Science Department, University of Crete

  2. ANNUAL SIZE OF THE DRAM SCALING GLOBAL DATASPHERE TREND 10000 200 MEGABITS/CHIP S S R R A ZETABYTES A 1000 E E 2X/3 YEARS 2X/3 YEARS Y Y 100 5 5 100 . . 1 1 / / X X 10 2 2 0 0 3 6 9 2 5 1 1 1 1 1 2 2 0 0 0 0 0 0 1985 1995 2005 2015 2 2 2 2 2 2 YEAR YEAR Data is growing faster while DRAM scaling is getting diffjcult 1

  3. ANNUAL SIZE OF THE NAND FLASH GLOBAL DATASPHERE SCALING TREND 200 4 DENSITY (TB) ZETABYTES 2 100 0 7 9 1 4 7 0 3 0 1 1 2 2 2 3 3 0 0 0 0 0 0 0 0 3 6 9 2 5 2 2 2 2 2 2 2 1 1 1 1 2 2 0 0 0 0 0 0 2 2 2 2 2 2 YEAR YEAR NAND Flash capacity is continuous scaling 2

  4. DATA-INTENSIVE APPLICATIONS DNA/PROTEIN DNA/PROTEIN VIRTUAL IMAGE VIRTUAL IMAGE SYNTHESIS SYNTHESIS REALITY ANALYSIS REALITY ANALYSIS IN-MEMORY FRAMEWORKS IN-MEMORY FRAMEWORKS More demand for memory More demand for memory 3

  5. APACHE SPARK IN-MEMORY COMPUTING RDD Operation 1 RDD RDD . . . RDD Operation n RDD RDD RDD RDD Operation 1 Operation n DISK RAM RAM DISK 4

  6. INTRODUCTION TO SPARK IN-MEMORY COMPUTING MEMORY_AND_DISK MEMORY_ONL Y SERIALIZE RDD RDD RDD RDD partition partition DISK MEMORY MEMORY MEMORY 5

  7. LET’S EXPLOIT THE CAPACITY OF STORAGE DEVICES JVM-based Analytics Frameworks Serialization / Memory-Mapped fjle I/O Deserialization We explore both approaches 6

  8. SERIALIZATION / DESERIALIZATION (LIMITATIONS) • Out-of-memory Errors due to small size of heaps. • Large computing results are generated during processing a record • Serialization / Deserialization afgects CPU performance • GC overhead to reclaim long-lived accumulated objects • Iterative applications 7

  9. ON-GOING WORK Non- DRAM JVM Device Heap Other Heap Heap fmap DRAM Storage Device 8

  10. ON-GOING WORK • Data placement policy inside JVM to manipulate Objects • Short-Lived data objects on DRAM Heap • Long-Lived data objects on Storage Device Heap • Add extra Storage Level in Apache Spark to support caching RDDs on Storage Heap • Thorough evaluation on SSDs, NVMe, Optane devices 9

  11. CONTACT INFORMATION Iacovos G. Kolokasis MSc Student, Computer Science Department, University of Crete kolokasis@ics.forth.gr Institute of Computer Science (ICS) Foundation for Research and Technology Hellas (FORTH) www.ics.forth.gr 10

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

DRAM Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays

DRAM Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays

DRAM Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays over time (us-scale) This is the dyanamic part. About 6F 2 : 20x better than SRAM Reading Precharge Assert word line

721 views • 32 slides
DRAM 1 Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays

DRAM 1 Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays

DRAM 1 Dynamic Random Access Memory (DRAM) Storage Charge on a capacitor Decays over time (us-scale) This is the dyanamic part. About 6F 2 : 20x better than SRAM Reading Precharge Assert word line

550 views • 31 slides
Module 6.1 Memory Access Performance DRAM Bandwidth Objective To learn that memory

Module 6.1 Memory Access Performance DRAM Bandwidth Objective To learn that memory

GPU Teaching Kit Accelerated Computing Module 6.1 Memory Access Performance DRAM Bandwidth Objective To learn that memory bandwidth is a first-order performance factor in a massively parallel processor DRAM bursts, banks, and

464 views • 12 slides
Large Scale DRAM Model DRAM Engineers DRAM Engineers Team: Abdulrahman Alqahtani,

Large Scale DRAM Model DRAM Engineers DRAM Engineers Team: Abdulrahman Alqahtani,

Large Scale DRAM Model DRAM Engineers DRAM Engineers Team: Abdulrahman Alqahtani, Demetria Shepherd, Colby Weber, Jinming Yang, Zeyu Zhang Client: Daniel Eichenberger, Microns Test Engineer/Recruiter Capstone Mentor: Ashwija

306 views • 8 slides
UMBC A B M A L T F O U M B C I M Y O R T 1 (12/8/04) I E S R C E O V U

UMBC A B M A L T F O U M B C I M Y O R T 1 (12/8/04) I E S R C E O V U

Advanced VLSI Design Memory CMPE 640 Memory Can be categorized into: Read Write Memory (RWM) Random Access Memory (RAM): static SRAM (faster) verses dynamic DRAM (smaller) structures possible. Access time independent of physical

381 views • 21 slides
Virtual Memory Lecture 25 CS301 DRAM as cache What about programs larger than DRAM?

Virtual Memory Lecture 25 CS301 DRAM as cache What about programs larger than DRAM?

Virtual Memory Lecture 25 CS301 DRAM as cache What about programs larger than DRAM? When we run multiple programs, all must fit in DRAM! Add another larger, slower level to the memory hierarchy - use part of the hard drive.

834 views • 61 slides
COMP 590-154: Computer Architecture Memory / DRAM SRAM vs. DRAM SRAM = Static RAM As

COMP 590-154: Computer Architecture Memory / DRAM SRAM vs. DRAM SRAM = Static RAM As

COMP 590-154: Computer Architecture Memory / DRAM SRAM vs. DRAM SRAM = Static RAM As long as power is present, data is retained DRAM = Dynamic RAM If you dont do anything, you lose the data SRAM: 6T per bit built with

1.4k views • 38 slides
Memory Hierarchy Instructor: Jun Yang 1 11/19/2009 Motivation Processor-DRAM Memory Gap

Memory Hierarchy Instructor: Jun Yang 1 11/19/2009 Motivation Processor-DRAM Memory Gap

Memory Hierarchy Instructor: Jun Yang 1 11/19/2009 Motivation Processor-DRAM Memory Gap (latency) Proc 1000 CPU 60%/yr. Moores Law (2X/1.5yr) Performance 100 Processor-Memory Performance Gap: (grows 50% / year) 10 DRAM

862 views • 36 slides
Chapter 5 Internal Memory Contents Semiconductor main memory Organization DRAM and

Chapter 5 Internal Memory Contents Semiconductor main memory Organization DRAM and

Chapter 5 Internal Memory Contents Semiconductor main memory Organization DRAM and SRAM Types of ROM Chip logic Chip packaging Module organization Error correction Advanced DRAM organization Synchronous

304 views • 27 slides
Emerging Non Volatile Memory Resistive Memory Technologies Key concept: replace DRAM cell

Emerging Non Volatile Memory Resistive Memory Technologies Key concept: replace DRAM cell

Emerging Non Volatile Memory Resistive Memory Technologies Key concept: replace DRAM cell capacitor with a programmable resistor 1T-1C DRAM 1T-1R STT-MRAM, PCM, RRAM Charge based sensing Resistance based sensing

666 views • 19 slides
Introduction Modern DRAM Memory Architectures Sam Miller Memory subsystem is a bottleneck

Introduction Modern DRAM Memory Architectures Sam Miller Memory subsystem is a bottleneck

Introduction Modern DRAM Memory Architectures Sam Miller Memory subsystem is a bottleneck Tam Chantem Memory stall time will become dominant Jon Lucas New architectures & accessing techniques proposed to combat these issues

422 views • 4 slides
DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah

DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah

DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah CS/ECE 6810: Computer Architecture Overview Announcement Homework 5 submission deadline: Apr. 10 th This lecture DRAM control DRAM

137 views • 11 slides
Main Memory & DRAM Nima Honarmand Spring 2018 :: CSE 502 Main Memory Big Picture 1)

Main Memory & DRAM Nima Honarmand Spring 2018 :: CSE 502 Main Memory Big Picture 1)

Spring 2018 :: CSE 502 Main Memory & DRAM Nima Honarmand Spring 2018 :: CSE 502 Main Memory Big Picture 1) Last-level cache sends its memory requests to a Memory Controller Over a system bus of other types of interconnect 2)

963 views • 41 slides
Who Cares About the Impact on Performance Memory Hierarchy? Suppose a processor executes at

Who Cares About the Impact on Performance Memory Hierarchy? Suppose a processor executes at

8/20/08 Technology Trends Computer System Architecture Capacity Speed (latency) Logic: 2x in 3 years 2x in 3 years DRAM: 4x in 3 years 2x in 10 years Memory Part I Disk: 4x in 3 years 2x in 10 years DRAM Chalermek

290 views • 5 slides
DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah

DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah

DRAM CONTROLLER Mahdi Nazm Bojnordi Assistant Professor School of Computing University of Utah CS/ECE 6810: Computer Architecture Overview Announcement Homework 4 will be released on Nov. 28 th This lecture DRAM control DRAM

654 views • 34 slides
Modern DRAM Memory Systems Brian T. Davis MTU Interview Seminar Advanced Computer Architecture

Modern DRAM Memory Systems Brian T. Davis MTU Interview Seminar Advanced Computer Architecture

Modern DRAM Memory Systems Brian T. Davis MTU Interview Seminar Advanced Computer Architecture Laboratory University of Michigan April 24, 2000 page 1 Brian T. Davis Introduction Memory system Research objective DRAM

525 views • 38 slides
Fitness + In-Memory Computing = Getting ahead of the game Craig Gresbrink Solutions Architect

Fitness + In-Memory Computing = Getting ahead of the game Craig Gresbrink Solutions Architect

Fitness + In-Memory Computing = Getting ahead of the game Craig Gresbrink Solutions Architect 24 Hour Fitness Who are we? 24 Hour Fitness is a leading fitness industry pioneer with more than 400 clubs across the United States. 24 Hour Fitness

355 views • 23 slides
The JVM is not observable enough (and what to do about it) Stephen Kell stephen.kell@usi.ch

The JVM is not observable enough (and what to do about it) Stephen Kell stephen.kell@usi.ch

The JVM is not observable enough (and what to do about it) Stephen Kell stephen.kell@usi.ch University of Lugano joint work with: Danilo Ansaloni, Walter Binder, Luk a s Marek The JVM is. . . p.1/20 0xcafebabe This is a talk

422 views • 23 slides
Pufferfish: Container-driven Elastic Memory Management for Data-intensive Applications Wei Chen,

Pufferfish: Container-driven Elastic Memory Management for Data-intensive Applications Wei Chen,

Pufferfish: Container-driven Elastic Memory Management for Data-intensive Applications Wei Chen, Aidi Pi , Shaoqi Wang and Xiaobo Zhou University of Colorado, Colorado Springs Outline Introduction to data-intensive applications Memory

336 views • 22 slides
Alibaba Dragonwell JDK: Towards a Java Runtime for Cloud Computing Xiaoming Gu Alibaba JVM Team

Alibaba Dragonwell JDK: Towards a Java Runtime for Cloud Computing Xiaoming Gu Alibaba JVM Team

Alibaba Dragonwell JDK: Towards a Java Runtime for Cloud Computing Xiaoming Gu Alibaba JVM Team 1 System Software (OS / JVM / Virtualization) Resource Scheduling / Cluster Management / Container Database / Storage / Middleware / Computing

2.11k views • 20 slides
An Experimental Evaluation of Garbage Collectors on Big Data Applications Lijie Xu 1 , Tian Guo 2 ,

An Experimental Evaluation of Garbage Collectors on Big Data Applications Lijie Xu 1 , Tian Guo 2 ,

The 45th International Conference on Very Large Data Bases (VLDB 2019) An Experimental Evaluation of Garbage Collectors on Big Data Applications Lijie Xu 1 , Tian Guo 2 , Wensheng Dou 1 , Wei Wang 1 , Jun Wei 1 1 Institute of Software, Chinese

456 views • 45 slides
JVM Implementation Challenges JVM Implementation Challenges: JVM Unit of execution is a class

JVM Implementation Challenges JVM Implementation Challenges: JVM Unit of execution is a class

Smalltalk on the JVM By James Ladd object@redline.st http://redline.st @redline_st - ESUG 2011 JVM Implementation Challenges JVM Implementation Challenges: JVM Unit of execution is a class Sequence of bytes in class format Class is not an

632 views • 42 slides
Numerical static analysis with Soot Gianluca Amato Universit` a G. dAnnunzio di

Numerical static analysis with Soot Gianluca Amato Universit` a G. dAnnunzio di

Numerical static analysis with Soot Gianluca Amato Universit` a G. dAnnunzio di ChietiPescara ACM SIGPLAN International Workshop on the State Of the Art in Java Program Analysis SOAP 2013 (joint work with Francesca Scozzari and

859 views • 62 slides
Link-Time Enforcement of Confined Types for JVM Bytecode Philip W. L. Fong pwlfong@cs.uregina.ca

Link-Time Enforcement of Confined Types for JVM Bytecode Philip W. L. Fong pwlfong@cs.uregina.ca

Link-Time Enforcement of Confined Types for JVM Bytecode Philip W. L. Fong pwlfong@cs.uregina.ca Department of Computer Science University of Regina Regina, Saskatchewan, Canada Overview Motivation Confined Types A Bytecode-level

348 views • 34 slides

More recommend