SlideShare ist ein Scribd-Unternehmen logo
1 von 34
Information Classification: General
CONTRIBUTE.
COLLABORATE.
COMMERCIALIZE.
December 8-10 | Virtual Event
Information Classification: General
December 8-10 | Virtual Event
Ziptilion™️: Boosting RISC-V with an Efficient and OS
Transparent Memory Compression System
Angelos Arelakis
Co-founder and CTO
ZeroPoint Technologies
#RISCVSUMMIT
Information Classification: General
About ZeroPoint Technologies Prof. Per Stenström
CSO, Co-founder/inventor
Internationally renowned memory architecture expert (ACM and
IEEE Fellow).
Professor at Chalmers University of Technology. Senior industry
experience from Sun Microsystems with a wide industry network.
Dr. Angelos Arelakis
CTO, Co-founder/inventor
Memory system architecture and ultra-fast data
compression expert.
Has received award from the King Carl XVI Gustaf’s
fund for science, technology and environment.
15 people strong R&D team:
• State of the Art Compression
Algorithm Research & Development
• Memory Architecture Optimization
• Linux Kernel Development
• High Volume ASIC/FPGA
Development and Manufacturing
The Main Memory Compression Invention
originates from Prof. Per Stenström in the
late 1990’s
Company founded 2016, by
Prof. Per Stenström and Dr. Angelos Arelakis
Real Time Memory Compression is a
Challenging problem and Hard to solve
ZeroPoint’s achievement is the result of:
• Decades of frontline research
• In-depth knowledge
• Unique mix of competencies
50 manyears invested in R&D
Information Classification: General
System performance scaling, as we knew it,
stopped due to a Moore’s Law and Dennard’s
Scaling decline.
Heterogenous Multi-Processor Systems
continue to deliver compute power, but the
memory hierarchy cannot keep up.
The computing system scaling challenge
40 years of Microprocessor Trend Data Image: Karl Rupp
Information Classification: General
System performance scaling, as we knew it,
stopped due to Moore’s Law and Dennard’s
Scaling decline.
Heterogenous Multi-Processor Systems
continue to deliver compute power, but the
memory hierarchy cannot keep up.
Applications demand ever increasing memory
capacity and bandwidth
The computing system scaling challenge
Information Classification: General
System performance scaling, as we knew it,
stopped due to Moore’s Law and Dennard’s
Scaling decline.
Heterogenous Multi-Processor Systems
continue to deliver compute power, but the
memory hierarchy cannot keep up.
Applications demand ever increasing memory
capacity and bandwidth
The computing system scaling challenge
Memory
compression can
bridge the gap
We need new and disruptive solutions
Information Classification: General
ZeroPoint’s product – Ziptilion™️
Value proposition
• ZeroPoint Technologies: The only real time memory compression
technology
• Ziptilion doubles the memory capacity
• Ziptilion doubles the effective memory bandwidth
Information Classification: General
CPU(1)
L2$
L1D$ L1I$
CPU(2)
L2$
L1D$ L1I$
CPU(3)
L2$
L1D$ L1I$
CPU(...)
L2$
L1D$ L1I$
DRAM ctrl
ZiptilionTM
DRAM ctrl
ZiptilionTM
CPU(n)
L2$
L1D$ L1I$
L3$
DIMM DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
GPU
Peripherals
I/O
Other
M a i n M e m o r y
C o m p r e s s i o n
Ready for Tape Out in
EuroEXA* (800MHz @28nm)
• General purpose and lossless
• 2-3x compression performance
• Fast – Real time
• Compact – Area efficient
• Strong patent portfolio (10 patent families)
and a pipeline of new innovations
*funded by European Union, Horizon2020 program
EuroEXA builds an HPC system
ZeroPoint’s product – Ziptilion™️
The only real time Memory Compression Technology
Information Classification: General
Ziptilion™️ IP building blocks and latency
• Read and write data paths are
pipelined to sustain high throughput
to/from memory
• #Compressors/#Decompressors
configurable based on the target
memory b/w requirements
• Address translation: TLB-like block
translates on-the-fly from the physical
to the compressed address space
CPU(1)
L2$
L1D$ L1I$
CPU(2)
L2$
L1D$ L1I$
CPU(3)
L2$
L1D$ L1I$
CPU(...)
L2$
L1D$ L1I$
DRAM ctrl
ZiptilionTM
DRAM ctrl
ZiptilionTM
CPU(n)
L2$
L1D$ L1I$
L3$
DIMM DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
DIMM
GPU
Peripherals
I/O
Other
Information Classification: General
Ziptilion™️ IP innovations
Low overhead compression-related metadata:
0.7-3% of memory depending on IP configuration
Efficient compression algorithms optimized to
minimize latency and increase compressibility
Hardware-accelerated and online memory and
traffic data analysis to tune compression
Hardware / Software synergy offers robust
management of the compressed memory
Information Classification: General
Ziptilion™️: Transparent integration to the
operating system (OS)
• Ziptilion IP is integrated in the memory access path
and compresses the data continuously
• Free memory space generated by Ziptilion IP
(hardware) is collected by the Ziptilion SW driver
• Ziptilion SW forms a page-oriented pseudo-RAM
device (VPC)
• VPC exposed to OS transparently by connecting the
Ziptilion SW driver to the Frontswap API
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Baseline uncompressed memory
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Baseline uncompressed memory
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Baseline uncompressed memory
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
State-of-art ZSWAP/ZRAM
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
State-of-art ZSWAP/ZRAM
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
State-of-art ZSWAP/ZRAM
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion™️ VPC vs state-of-art vs baseline
Ziptilion VPC
Information Classification: General
Ziptilion: Compression algorithms implemented in
hardware today
Ziptilion+: Next IP generation
1
1.5
2
2.5
3
3.5
4
Compression
Ratio
(times)
SPEC2017
Memory compressibility with Ziptilion and Ziptilion+
Ziptilion
Ziptilion+
Compression ratio for SPEC2017
SPEC2017INT SPEC2017FP
• Ziptilion: Compresses SPEC2017 on average by
2.2x
• Ziptilion+: More applications are compressed by >2x
23.4
27.6
Ziptilion IP offers important compression opportunities: 2.2-2.6x
Information Classification: General
Ziptilion: Compression algorithms implemented in
hardware today
Ziptilion+: Next IP generation
Compression ratio for MLPerf/Training
Promising compression for MLPerf/Training workloads: 2.5x-3x
Information Classification: General
• No Ziptilion / 1x memory: Baseline system
without Ziptilion. Memory is adjusted to fit 50% of
the application’s memory footprint
• Results are normalized to this baseline
• Ziptilion / 1x memory: Above baseline system
enhanced with Ziptilion
• No Ziptilion / 2x memory: Baseline system
without Ziptilion and with enough memory to fit
the application footprint in memory
Ziptilion™️ performance for SPEC2017
Ziptilion offers 20% higher performance
Performance improvement close to a system with 2x more physical memory
Information Classification: General
ZSWAP+/ZRAM+: ZeroPoint hardware
accelerated ZSWAP/ZRAM
Baseline ZSWAP/ZRAM swap out
Baseline ZSWAP/ZRAM swap in
Information Classification: General
ZSWAP+/ZRAM+: ZeroPoint hardware
accelerated ZSWAP/ZRAM
• Lightweight Ziptilion IP containing mainly
compression and decompression accelerators
integrated in the memory access path
• ZSWAP/ZRAM API is modified to call the
hardware accelerators instead of SW-based
compression
Operation ZSWAP+/ZRAM+
Compression acceleration 15x
Decompression acceleration 10x
10-15x faster ZSWAP/ZRAM compression & decompression
crypto_compress_(a)comp
API
HW accelerated
ZComp
HW accelerated
ZDecomp
crypto_compress_(a)comp
API
Information Classification: General
ZSWAP+/ZRAM+
• HW accelerated ZSWAP/ZRAM
• Superfast, low power ZRAM/ZSWAP device
• Supporting Android and Linux based platforms (Smart
devices, Laptops and Servers) that benefit from a hardware
accelerated compressed swap device
ZSWAP+/ZRAM+
Entry version
Speed: 1.4x speed-up
License: Free of charge License
ZSWAP+/ZRAM+
Performance version
Speed: 2x speed-up
License: Paid License
ZSWAP+/ZRAM+: ZeroPoint hardware
accelerated ZSWAP/ZRAM
Available
Q1-2021
Contact ZeroPoint for evaluation!
Information Classification: General
Compression+Encryption
The cost of security is negligible when combined with compression
Standard encryption algorithms such as AES-XTS and new ciphers such as PRINCE
Information Classification: General
• Mature product –
Transparent integration to OS and Application
• Typically IP size: ~0.4mm2 @7nm
• High Performance and High Security –
Combining Compression and Encryption
Server CPU customer
“DRAM is super expensive!”
50% increased Performance / Watt
Performance:
Bandwidth:
Capacity:
20-50% increased System
Performance
Up to 50% increased available
Memory Bandwidth
2-3x increased Memory Capacity
ZeroPoint technology delivers
significant customer value
Information Classification: General
December 8-10 | Virtual Event
Thank you for joining us.
Contribute to the RISC-V conversation on social!
#RISCVSUMMIT @risc_v

Weitere ähnliche Inhalte

Was ist angesagt?

OSC2011 Tokyo/Fall 濃いバナ(virtio)
OSC2011 Tokyo/Fall 濃いバナ(virtio)OSC2011 Tokyo/Fall 濃いバナ(virtio)
OSC2011 Tokyo/Fall 濃いバナ(virtio)
Takeshi HASEGAWA
 
/proc/irq/<irq>/smp_affinity
/proc/irq/<irq>/smp_affinity/proc/irq/<irq>/smp_affinity
/proc/irq/<irq>/smp_affinity
Takuya ASADA
 

Was ist angesagt? (20)

DPDK in Containers Hands-on Lab
DPDK in Containers Hands-on LabDPDK in Containers Hands-on Lab
DPDK in Containers Hands-on Lab
 
Getting Started with Buildroot
Getting Started with BuildrootGetting Started with Buildroot
Getting Started with Buildroot
 
Intel dpdk Tutorial
Intel dpdk TutorialIntel dpdk Tutorial
Intel dpdk Tutorial
 
ARMアーキテクチャにおけるセキュリティ機構の紹介
ARMアーキテクチャにおけるセキュリティ機構の紹介ARMアーキテクチャにおけるセキュリティ機構の紹介
ARMアーキテクチャにおけるセキュリティ機構の紹介
 
Ixgbe internals
Ixgbe internalsIxgbe internals
Ixgbe internals
 
Build your own embedded linux distributions by yocto project
Build your own embedded linux distributions by yocto projectBuild your own embedded linux distributions by yocto project
Build your own embedded linux distributions by yocto project
 
Benefits of Multi-rail Cluster Architectures for GPU-based Nodes
Benefits of Multi-rail Cluster Architectures for GPU-based NodesBenefits of Multi-rail Cluster Architectures for GPU-based Nodes
Benefits of Multi-rail Cluster Architectures for GPU-based Nodes
 
Effective Linux Development Using PetaLinux Tools 2017.4
Effective Linux Development Using PetaLinux Tools 2017.4Effective Linux Development Using PetaLinux Tools 2017.4
Effective Linux Development Using PetaLinux Tools 2017.4
 
InnoDB Locking Explained with Stick Figures
InnoDB Locking Explained with Stick FiguresInnoDB Locking Explained with Stick Figures
InnoDB Locking Explained with Stick Figures
 
x86とコンテキストスイッチ
x86とコンテキストスイッチx86とコンテキストスイッチ
x86とコンテキストスイッチ
 
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Way Improved :) GC Tuning Confessions - presented at JavaOne2015Way Improved :) GC Tuning Confessions - presented at JavaOne2015
Way Improved :) GC Tuning Confessions - presented at JavaOne2015
 
Build a Deep Learning Video Analytics Framework | SIGGRAPH 2019 Technical Ses...
Build a Deep Learning Video Analytics Framework | SIGGRAPH 2019 Technical Ses...Build a Deep Learning Video Analytics Framework | SIGGRAPH 2019 Technical Ses...
Build a Deep Learning Video Analytics Framework | SIGGRAPH 2019 Technical Ses...
 
Hardware accelerated Virtualization in the ARM Cortex™ Processors
Hardware accelerated Virtualization in the ARM Cortex™ ProcessorsHardware accelerated Virtualization in the ARM Cortex™ Processors
Hardware accelerated Virtualization in the ARM Cortex™ Processors
 
OSC2011 Tokyo/Fall 濃いバナ(virtio)
OSC2011 Tokyo/Fall 濃いバナ(virtio)OSC2011 Tokyo/Fall 濃いバナ(virtio)
OSC2011 Tokyo/Fall 濃いバナ(virtio)
 
IDA ユーザなら知っておくべきマントノン侯爵夫人にモテる 7つの法則
IDA ユーザなら知っておくべきマントノン侯爵夫人にモテる 7つの法則IDA ユーザなら知っておくべきマントノン侯爵夫人にモテる 7つの法則
IDA ユーザなら知っておくべきマントノン侯爵夫人にモテる 7つの法則
 
Marco Cavallini - Yocto Project, an automatic generator of embedded Linux dis...
Marco Cavallini - Yocto Project, an automatic generator of embedded Linux dis...Marco Cavallini - Yocto Project, an automatic generator of embedded Linux dis...
Marco Cavallini - Yocto Project, an automatic generator of embedded Linux dis...
 
/proc/irq/<irq>/smp_affinity
/proc/irq/<irq>/smp_affinity/proc/irq/<irq>/smp_affinity
/proc/irq/<irq>/smp_affinity
 
Vivado hls勉強会1(基礎編)
Vivado hls勉強会1(基礎編)Vivado hls勉強会1(基礎編)
Vivado hls勉強会1(基礎編)
 
[JPCERT/CC POC Meeting] 研究紹介 + DLLハイジャックの脆弱性
[JPCERT/CC POC Meeting] 研究紹介 + DLLハイジャックの脆弱性[JPCERT/CC POC Meeting] 研究紹介 + DLLハイジャックの脆弱性
[JPCERT/CC POC Meeting] 研究紹介 + DLLハイジャックの脆弱性
 
Zynq VIPを利用したテストベンチ
Zynq VIPを利用したテストベンチZynq VIPを利用したテストベンチ
Zynq VIPを利用したテストベンチ
 

Ähnlich wie Ziptillion boosting RISC-V with an efficient and os transparent memory compression system

“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
Edge AI and Vision Alliance
 
Spectrum Scale final
Spectrum Scale finalSpectrum Scale final
Spectrum Scale final
Joe Krotz
 

Ähnlich wie Ziptillion boosting RISC-V with an efficient and os transparent memory compression system (20)

Ceph on Intel: Intel Storage Components, Benchmarks, and Contributions
Ceph on Intel: Intel Storage Components, Benchmarks, and ContributionsCeph on Intel: Intel Storage Components, Benchmarks, and Contributions
Ceph on Intel: Intel Storage Components, Benchmarks, and Contributions
 
Ceph on Intel: Intel Storage Components, Benchmarks, and Contributions
Ceph on Intel: Intel Storage Components, Benchmarks, and ContributionsCeph on Intel: Intel Storage Components, Benchmarks, and Contributions
Ceph on Intel: Intel Storage Components, Benchmarks, and Contributions
 
How do you pick the right Storage vendor?
How do you pick the right Storage vendor?How do you pick the right Storage vendor?
How do you pick the right Storage vendor?
 
Increase Your Mission Critical Application Performance without Breaking the B...
Increase Your Mission Critical Application Performance without Breaking the B...Increase Your Mission Critical Application Performance without Breaking the B...
Increase Your Mission Critical Application Performance without Breaking the B...
 
Webinar NETGEAR - Storagecraft e Netgear: soluzioni per il backup e il disast...
Webinar NETGEAR - Storagecraft e Netgear: soluzioni per il backup e il disast...Webinar NETGEAR - Storagecraft e Netgear: soluzioni per il backup e il disast...
Webinar NETGEAR - Storagecraft e Netgear: soluzioni per il backup e il disast...
 
Ceph Day Beijing - Storage Modernization with Intel & Ceph
Ceph Day Beijing - Storage Modernization with Intel & Ceph Ceph Day Beijing - Storage Modernization with Intel & Ceph
Ceph Day Beijing - Storage Modernization with Intel & Ceph
 
Ceph Day Beijing - Storage Modernization with Intel and Ceph
Ceph Day Beijing - Storage Modernization with Intel and CephCeph Day Beijing - Storage Modernization with Intel and Ceph
Ceph Day Beijing - Storage Modernization with Intel and Ceph
 
Emc isilon technical deep dive workshop
Emc isilon technical deep dive workshopEmc isilon technical deep dive workshop
Emc isilon technical deep dive workshop
 
Emc isilon overview
Emc isilon overview Emc isilon overview
Emc isilon overview
 
Live CEO Interview and Webinar Update on the State of Deduplication
 Live CEO Interview and Webinar Update on the State of Deduplication Live CEO Interview and Webinar Update on the State of Deduplication
Live CEO Interview and Webinar Update on the State of Deduplication
 
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
“Using a Neural Processor for Always-sensing Cameras,” a Presentation from Ex...
 
Innodisk at aditech customer meet 2015
Innodisk at aditech customer meet 2015Innodisk at aditech customer meet 2015
Innodisk at aditech customer meet 2015
 
Presentazione SimpliVity @ VMUGIT UserCon 2015
Presentazione SimpliVity @ VMUGIT UserCon 2015Presentazione SimpliVity @ VMUGIT UserCon 2015
Presentazione SimpliVity @ VMUGIT UserCon 2015
 
Nervana and the Future of Computing
Nervana and the Future of ComputingNervana and the Future of Computing
Nervana and the Future of Computing
 
HP flash optimized storage - webcast
HP flash optimized storage - webcastHP flash optimized storage - webcast
HP flash optimized storage - webcast
 
DataCore Software with Cisco UCS Complete Unification of the Data Center Ser...
 DataCore Software with Cisco UCS Complete Unification of the Data Center Ser... DataCore Software with Cisco UCS Complete Unification of the Data Center Ser...
DataCore Software with Cisco UCS Complete Unification of the Data Center Ser...
 
Zenith Infotech Mirror Cloud Presentation. 112211
Zenith Infotech    Mirror Cloud Presentation. 112211Zenith Infotech    Mirror Cloud Presentation. 112211
Zenith Infotech Mirror Cloud Presentation. 112211
 
Spectrum Scale final
Spectrum Scale finalSpectrum Scale final
Spectrum Scale final
 
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage ComparisonIntel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
Intel and DataStax: 3D XPoint and NVME Technology Cassandra Storage Comparison
 
Garance 100% dostupnosti dat! Kdo z vás to má?
Garance 100% dostupnosti dat! Kdo z vás to má?Garance 100% dostupnosti dat! Kdo z vás to má?
Garance 100% dostupnosti dat! Kdo z vás to má?
 

Mehr von RISC-V International

Mehr von RISC-V International (20)

WD RISC-V inliner work effort
WD RISC-V inliner work effortWD RISC-V inliner work effort
WD RISC-V inliner work effort
 
RISC-V Zce Extension
RISC-V Zce ExtensionRISC-V Zce Extension
RISC-V Zce Extension
 
RISC-V Online Tutor
RISC-V Online TutorRISC-V Online Tutor
RISC-V Online Tutor
 
London Open Source Meetup for RISC-V
London Open Source Meetup for RISC-VLondon Open Source Meetup for RISC-V
London Open Source Meetup for RISC-V
 
RISC-V Introduction
RISC-V IntroductionRISC-V Introduction
RISC-V Introduction
 
Static partitioning virtualization on RISC-V
Static partitioning virtualization on RISC-VStatic partitioning virtualization on RISC-V
Static partitioning virtualization on RISC-V
 
Standardizing the tee with global platform and RISC-V
Standardizing the tee with global platform and RISC-VStandardizing the tee with global platform and RISC-V
Standardizing the tee with global platform and RISC-V
 
Semi dynamics high bandwidth vector capable RISC-V cores
Semi dynamics high bandwidth vector capable RISC-V coresSemi dynamics high bandwidth vector capable RISC-V cores
Semi dynamics high bandwidth vector capable RISC-V cores
 
Security and functional safety
Security and functional safetySecurity and functional safety
Security and functional safety
 
Reverse Engineering of Rocket Chip
Reverse Engineering of Rocket ChipReverse Engineering of Rocket Chip
Reverse Engineering of Rocket Chip
 
RISC-V NOEL-V - A new high performance RISC-V Processor Family
RISC-V NOEL-V - A new high performance RISC-V Processor FamilyRISC-V NOEL-V - A new high performance RISC-V Processor Family
RISC-V NOEL-V - A new high performance RISC-V Processor Family
 
RISC-V 30910 kassem_ summit 2020 - so_c_gen
RISC-V 30910 kassem_ summit 2020 - so_c_genRISC-V 30910 kassem_ summit 2020 - so_c_gen
RISC-V 30910 kassem_ summit 2020 - so_c_gen
 
RISC-V 30908 patra
RISC-V 30908 patraRISC-V 30908 patra
RISC-V 30908 patra
 
RISC-V 30907 summit 2020 joint picocom_mentor
RISC-V 30907 summit 2020 joint picocom_mentorRISC-V 30907 summit 2020 joint picocom_mentor
RISC-V 30907 summit 2020 joint picocom_mentor
 
RISC-V 30906 hex five multi_zone iot firmware
RISC-V 30906 hex five multi_zone iot firmwareRISC-V 30906 hex five multi_zone iot firmware
RISC-V 30906 hex five multi_zone iot firmware
 
RISC-V 30946 manuel_offenberg_v3_notes
RISC-V 30946 manuel_offenberg_v3_notesRISC-V 30946 manuel_offenberg_v3_notes
RISC-V 30946 manuel_offenberg_v3_notes
 
RISC-V software state of the union
RISC-V software state of the unionRISC-V software state of the union
RISC-V software state of the union
 
Ripes tracking computer architecture throught visual and interactive simula...
Ripes   tracking computer architecture throught visual and interactive simula...Ripes   tracking computer architecture throught visual and interactive simula...
Ripes tracking computer architecture throught visual and interactive simula...
 
Porting tock to open titan
Porting tock to open titanPorting tock to open titan
Porting tock to open titan
 
Open j9 jdk on RISC-V
Open j9 jdk on RISC-VOpen j9 jdk on RISC-V
Open j9 jdk on RISC-V
 

Kürzlich hochgeladen

Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
vu2urc
 

Kürzlich hochgeladen (20)

Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your Business
 

Ziptillion boosting RISC-V with an efficient and os transparent memory compression system

  • 2. Information Classification: General December 8-10 | Virtual Event Ziptilion™️: Boosting RISC-V with an Efficient and OS Transparent Memory Compression System Angelos Arelakis Co-founder and CTO ZeroPoint Technologies #RISCVSUMMIT
  • 3. Information Classification: General About ZeroPoint Technologies Prof. Per Stenström CSO, Co-founder/inventor Internationally renowned memory architecture expert (ACM and IEEE Fellow). Professor at Chalmers University of Technology. Senior industry experience from Sun Microsystems with a wide industry network. Dr. Angelos Arelakis CTO, Co-founder/inventor Memory system architecture and ultra-fast data compression expert. Has received award from the King Carl XVI Gustaf’s fund for science, technology and environment. 15 people strong R&D team: • State of the Art Compression Algorithm Research & Development • Memory Architecture Optimization • Linux Kernel Development • High Volume ASIC/FPGA Development and Manufacturing The Main Memory Compression Invention originates from Prof. Per Stenström in the late 1990’s Company founded 2016, by Prof. Per Stenström and Dr. Angelos Arelakis Real Time Memory Compression is a Challenging problem and Hard to solve ZeroPoint’s achievement is the result of: • Decades of frontline research • In-depth knowledge • Unique mix of competencies 50 manyears invested in R&D
  • 4. Information Classification: General System performance scaling, as we knew it, stopped due to a Moore’s Law and Dennard’s Scaling decline. Heterogenous Multi-Processor Systems continue to deliver compute power, but the memory hierarchy cannot keep up. The computing system scaling challenge 40 years of Microprocessor Trend Data Image: Karl Rupp
  • 5. Information Classification: General System performance scaling, as we knew it, stopped due to Moore’s Law and Dennard’s Scaling decline. Heterogenous Multi-Processor Systems continue to deliver compute power, but the memory hierarchy cannot keep up. Applications demand ever increasing memory capacity and bandwidth The computing system scaling challenge
  • 6. Information Classification: General System performance scaling, as we knew it, stopped due to Moore’s Law and Dennard’s Scaling decline. Heterogenous Multi-Processor Systems continue to deliver compute power, but the memory hierarchy cannot keep up. Applications demand ever increasing memory capacity and bandwidth The computing system scaling challenge Memory compression can bridge the gap We need new and disruptive solutions
  • 7. Information Classification: General ZeroPoint’s product – Ziptilion™️ Value proposition • ZeroPoint Technologies: The only real time memory compression technology • Ziptilion doubles the memory capacity • Ziptilion doubles the effective memory bandwidth
  • 8. Information Classification: General CPU(1) L2$ L1D$ L1I$ CPU(2) L2$ L1D$ L1I$ CPU(3) L2$ L1D$ L1I$ CPU(...) L2$ L1D$ L1I$ DRAM ctrl ZiptilionTM DRAM ctrl ZiptilionTM CPU(n) L2$ L1D$ L1I$ L3$ DIMM DIMM DIMM DIMM DIMM DIMM DIMM DIMM GPU Peripherals I/O Other M a i n M e m o r y C o m p r e s s i o n Ready for Tape Out in EuroEXA* (800MHz @28nm) • General purpose and lossless • 2-3x compression performance • Fast – Real time • Compact – Area efficient • Strong patent portfolio (10 patent families) and a pipeline of new innovations *funded by European Union, Horizon2020 program EuroEXA builds an HPC system ZeroPoint’s product – Ziptilion™️ The only real time Memory Compression Technology
  • 9. Information Classification: General Ziptilion™️ IP building blocks and latency • Read and write data paths are pipelined to sustain high throughput to/from memory • #Compressors/#Decompressors configurable based on the target memory b/w requirements • Address translation: TLB-like block translates on-the-fly from the physical to the compressed address space CPU(1) L2$ L1D$ L1I$ CPU(2) L2$ L1D$ L1I$ CPU(3) L2$ L1D$ L1I$ CPU(...) L2$ L1D$ L1I$ DRAM ctrl ZiptilionTM DRAM ctrl ZiptilionTM CPU(n) L2$ L1D$ L1I$ L3$ DIMM DIMM DIMM DIMM DIMM DIMM DIMM DIMM GPU Peripherals I/O Other
  • 10. Information Classification: General Ziptilion™️ IP innovations Low overhead compression-related metadata: 0.7-3% of memory depending on IP configuration Efficient compression algorithms optimized to minimize latency and increase compressibility Hardware-accelerated and online memory and traffic data analysis to tune compression Hardware / Software synergy offers robust management of the compressed memory
  • 11. Information Classification: General Ziptilion™️: Transparent integration to the operating system (OS) • Ziptilion IP is integrated in the memory access path and compresses the data continuously • Free memory space generated by Ziptilion IP (hardware) is collected by the Ziptilion SW driver • Ziptilion SW forms a page-oriented pseudo-RAM device (VPC) • VPC exposed to OS transparently by connecting the Ziptilion SW driver to the Frontswap API
  • 12. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Baseline uncompressed memory
  • 13. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Baseline uncompressed memory
  • 14. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Baseline uncompressed memory
  • 15. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline State-of-art ZSWAP/ZRAM
  • 16. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline State-of-art ZSWAP/ZRAM
  • 17. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline State-of-art ZSWAP/ZRAM
  • 18. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 19. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 20. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 21. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 22. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 23. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 24. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 25. Information Classification: General Ziptilion™️ VPC vs state-of-art vs baseline Ziptilion VPC
  • 26. Information Classification: General Ziptilion: Compression algorithms implemented in hardware today Ziptilion+: Next IP generation 1 1.5 2 2.5 3 3.5 4 Compression Ratio (times) SPEC2017 Memory compressibility with Ziptilion and Ziptilion+ Ziptilion Ziptilion+ Compression ratio for SPEC2017 SPEC2017INT SPEC2017FP • Ziptilion: Compresses SPEC2017 on average by 2.2x • Ziptilion+: More applications are compressed by >2x 23.4 27.6 Ziptilion IP offers important compression opportunities: 2.2-2.6x
  • 27. Information Classification: General Ziptilion: Compression algorithms implemented in hardware today Ziptilion+: Next IP generation Compression ratio for MLPerf/Training Promising compression for MLPerf/Training workloads: 2.5x-3x
  • 28. Information Classification: General • No Ziptilion / 1x memory: Baseline system without Ziptilion. Memory is adjusted to fit 50% of the application’s memory footprint • Results are normalized to this baseline • Ziptilion / 1x memory: Above baseline system enhanced with Ziptilion • No Ziptilion / 2x memory: Baseline system without Ziptilion and with enough memory to fit the application footprint in memory Ziptilion™️ performance for SPEC2017 Ziptilion offers 20% higher performance Performance improvement close to a system with 2x more physical memory
  • 29. Information Classification: General ZSWAP+/ZRAM+: ZeroPoint hardware accelerated ZSWAP/ZRAM Baseline ZSWAP/ZRAM swap out Baseline ZSWAP/ZRAM swap in
  • 30. Information Classification: General ZSWAP+/ZRAM+: ZeroPoint hardware accelerated ZSWAP/ZRAM • Lightweight Ziptilion IP containing mainly compression and decompression accelerators integrated in the memory access path • ZSWAP/ZRAM API is modified to call the hardware accelerators instead of SW-based compression Operation ZSWAP+/ZRAM+ Compression acceleration 15x Decompression acceleration 10x 10-15x faster ZSWAP/ZRAM compression & decompression crypto_compress_(a)comp API HW accelerated ZComp HW accelerated ZDecomp crypto_compress_(a)comp API
  • 31. Information Classification: General ZSWAP+/ZRAM+ • HW accelerated ZSWAP/ZRAM • Superfast, low power ZRAM/ZSWAP device • Supporting Android and Linux based platforms (Smart devices, Laptops and Servers) that benefit from a hardware accelerated compressed swap device ZSWAP+/ZRAM+ Entry version Speed: 1.4x speed-up License: Free of charge License ZSWAP+/ZRAM+ Performance version Speed: 2x speed-up License: Paid License ZSWAP+/ZRAM+: ZeroPoint hardware accelerated ZSWAP/ZRAM Available Q1-2021 Contact ZeroPoint for evaluation!
  • 32. Information Classification: General Compression+Encryption The cost of security is negligible when combined with compression Standard encryption algorithms such as AES-XTS and new ciphers such as PRINCE
  • 33. Information Classification: General • Mature product – Transparent integration to OS and Application • Typically IP size: ~0.4mm2 @7nm • High Performance and High Security – Combining Compression and Encryption Server CPU customer “DRAM is super expensive!” 50% increased Performance / Watt Performance: Bandwidth: Capacity: 20-50% increased System Performance Up to 50% increased available Memory Bandwidth 2-3x increased Memory Capacity ZeroPoint technology delivers significant customer value
  • 34. Information Classification: General December 8-10 | Virtual Event Thank you for joining us. Contribute to the RISC-V conversation on social! #RISCVSUMMIT @risc_v