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

▶
in memory computing at scale look beyond physical dram
SMART_READER_LITE
LIVE PREVIEW

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

Jan 24, 2024 317 likes •446 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

slide-1
SLIDE 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

slide-2
SLIDE 2

ANNUAL SIZE OF THE GLOBAL DATASPHERE

1

Data is growing faster while DRAM scaling is getting diffjcult 1985 1995 2005 2015 1 10 100 1000 10000 YEAR MEGABITS/CHIP

2 X / 1 . 5 Y E A R S 2 X / 1 . 5 Y E A R S 2X/3 YEARS 2X/3 YEARS

DRAM SCALING TREND 2 1 2 1 3 2 1 6 2 1 9 2 2 2 2 2 5 100 200 YEAR ZETABYTES

slide-3
SLIDE 3

ANNUAL SIZE OF THE GLOBAL DATASPHERE

2

NAND Flash capacity is continuous scaling 2 1 7 2 1 9 2 2 1 2 2 4 2 2 7 2 3 2 3 3 2 4 YEAR DENSITY (TB) NAND FLASH SCALING TREND 2 1 2 1 3 2 1 6 2 1 9 2 2 2 2 2 5 100 200 YEAR ZETABYTES

slide-4
SLIDE 4

DATA-INTENSIVE APPLICATIONS More demand for memory More demand for memory

3

DNA/PROTEIN SYNTHESIS DNA/PROTEIN SYNTHESIS IN-MEMORY FRAMEWORKS IN-MEMORY FRAMEWORKS IMAGE ANALYSIS IMAGE ANALYSIS VIRTUAL REALITY VIRTUAL REALITY

slide-5
SLIDE 5

APACHE SPARK IN-MEMORY COMPUTING

4

DISK RAM Operation 1 RAM Operation n DISK Operation 1 Operation n . . . RDD RDD RDD RDD RDD RDD RDD RDD

slide-6
SLIDE 6

5

INTRODUCTION TO SPARK IN-MEMORY COMPUTING

RDD MEMORY MEMORY

MEMORY_ONL Y

RDD DISK MEMORY

MEMORY_AND_DISK

SERIALIZE RDD partition RDD partition

slide-7
SLIDE 7

LET’S EXPLOIT THE CAPACITY OF STORAGE DEVICES

6

Serialization / Deserialization Memory-Mapped fjle I/O

We explore both approaches

JVM-based Analytics Frameworks

slide-8
SLIDE 8

7

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
slide-9
SLIDE 9

8

ON-GOING WORK

DRAM Heap Non- Heap Other DRAM Storage Device fmap Device Heap JVM

slide-10
SLIDE 10
  • 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
  • n Storage Heap
  • Thorough evaluation on SSDs, NVMe, Optane devices

9

ON-GOING WORK

slide-11
SLIDE 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