SlideShare a Scribd company logo
1 of 22
Three Reasons Why NAS is No Good
for AI and Machine Learning
● NAS can’t leverage today’s flash technology
● NAS has no or very rudimentary cloud integration
● NAS data protection schemes are expensive
Watch and learn why:
For audio playback and Q&A go to: bit.ly/NASAIML
OurSpeakers
Sean Kerr, Product
Manager - HPC/AI Storage
Solutions, HPE
George Crump,
Founder and Lead Analyst of
Storage Switzerland
Barbara Murphy,
VP of Marketing,
WekaIO
What is
Artificial Intelligence (AI)
and Machine Learning (ML)?
Top Mainstream Use
Cases for AI and ML
● Autonomous Vehicles
● Fraud and Theft Detection
● Customer Service (ChatBots)
● Logistic Management
● Cyber-Security
● Smart Cities
The Data
Requirements for
AI and ML
● Extreme High-Performance IO
● Low Latency
● Adept at High File Count
● Massive Scalability
(Capacity and Performance)
● Multi-Location Processing
(Multiple Data Centers and Cloud)
AI and ML
Tends to
“Sneak Up”
on Core IT
● Starts as a pilot or skunk-works project
● Uses existing storage resources or cloud
storage
● Moves into production, breaking
traditional storage
● IT tries to upgrade/expand traditional
NAS
○ The project gets costly and can’t meet
requirements
3 Reasons Why NAS
Won’t Work for AI/ML
NAS is a 90’S
Technology
● Not architected to
leverage flash
● Compute architecture has
changed (GPUs + CPUs)
● NAS has to deliver high
performance
NAS Doesn’t Know Cloud
● Today’s NAS (file systems) have rudimentary cloud
support
○ Typically replication for DR
● AI/ML Requires
○ Successful AI outcomes require keeping the
compute layer saturated with data
● Seamless movement of data to the cloud for
processing
○ Recall data (the results) from the cloud for
processing
○ Archive data to the cloud for long-term storage
NAS Data
Protection is
Expensive
● Impacts performance
○ Especially during failed-state
● Costly
○ Requires too much resource overhead
(compute and capacity)
● Slow
○ Return to good-state requires either time
or a lot of CPU
● Software not Hardware
● High Performance
○ Leverage Flash
○ Leverage NVMe
IT Needs a Modern Approach
● More of a Fabric than a File
System
○ Data moves seamlessly
between locations and cloud
○ Exploits cloud tiers and
compute
● Data Protection Designed for
AI/ML
11© 2018 All rights reserved.
How WekaIO Solves the AI
Challenge
Barbara Murphy
VP of Marketing
July 25, 2018 | 9:00 a.m. PDT
12© 2018 All rights reserved.
NFS = Not For Speed
GPU Bandwidth
11GBytes/sec/IB
connection
NFS Bandwidth
1 - 1.5GBytes/sec
o A protocol invented in 1984 trying to solve a 2018 problem
o pNFS tried to fix NFS but failed when metadata workloads exploded
o Legacy parallel file systems like Lustre and GPFS cannot handle billions of
small files
– And they are complex to operate
13© 2018 All rights reserved.
The Power of Parallelism
o Modern networks on 100Gbit Ethernet are 100x faster than SSD
o Everything can be distributed
o With NVMe-oF, shared storage is faster than local storage
14© 2018 All rights reserved.
WekaIO Solves the Data Accessibility Problem
o Shared, POSIX compliant Parallel file system written for NVMe
o Scales to trillions of files, billions in a single directory
o Simple to install and use
GPU Bandwidth
4-11GBytes/sec
Per GPU Server
11GBytes/sec
15© 2018 All rights reserved.
WekaIO Matrix: Full-featured and Flexible
WekaIO Matrix Shared File System
Fully Coherent POSIX File System That is Faster than Local File System
Distributed Coding, More Resilient at Scale, Fast Rebuilds, End-to-End DP
Instantaneous Snapshots, Clones, Tiering to S3, Partial File Rehydration
InfiniBand or Ethernet, Converged or Dedicated Storage Server
HPE AWS Cloud
16© 2018 All rights reserved.
WekaIO is Faster than Local NVMe Drives
Testing conducted on WekaIO/HPE reference platform
• Inference benchmarks are I/O Bound
• Single Client Performance on HPE Apollo 6500
• 8 NVIDIA V100 GPUs
• WekaIO Matrix on 8x HPE DL360
• Local NVMe FS is I/O Bound
http://dlpg.labs.hpe.com report 10 and 11
17© 2018 All rights reserved.
Results From Deep Learning Use Case
Actual measured data to the AI/ML GPU Cluster
WekaIO is over 10x Faster than NFS All Flash
NAS
Initial tests
Production
HPE & WekaIO
an Engineered Solution
Sean Kerr
WekaIO Matrix – Solution Building Block
19a00045311enw
Apollo 2000 Gen10
ProLiant DL360
Gen10
Supported Storage Systems for
WekaIO Matrix
Apollo 6500 Gen10
Mellanox Switch
WekaIO with
Apollo 2000
Thank you!
Storage Switzerland
http://www.storageswiss.com
georgeacrump@storageswiss.com
StorageSwiss on Twitter:
http://twitter.com/storageswiss
StorageSwiss on YouTube:
http://www.youtube.com/user/storageswiss
WekaIO
https://www.weka.io
info@weka.io
WekaIO on Twitter: @Wekaio
Barbara Murphy on Twitter: @scaleoutlady
WekaIO on LinkedIN:
https://www.linkedin.com/company/weka-io
HPE
https://www.hpe.com
HPE on Twitter:
https://twitter.com/hpe_hpc
https://twitter.com/HPE_Servers
HPE on Facebook:
https://www.facebook.com/HPEServers/
Three Reasons Why NAS is No Good for AI and Machine Learning
For complete audio and Q&A please register for the On Demand Version:
bit.ly/NASAIML

More Related Content

What's hot

Realtime Analytical Query Processing and Predictive Model Building on High Di...
Realtime Analytical Query Processing and Predictive Model Building on High Di...Realtime Analytical Query Processing and Predictive Model Building on High Di...
Realtime Analytical Query Processing and Predictive Model Building on High Di...
Spark Summit
 
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
Redis Labs
 
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
Alluxio, Inc.
 
cleversafe_definitive_guide_white_paper
cleversafe_definitive_guide_white_papercleversafe_definitive_guide_white_paper
cleversafe_definitive_guide_white_paper
Chris Woeppel
 

What's hot (20)

Accelerating Data Computation on Ceph Objects
Accelerating Data Computation on Ceph ObjectsAccelerating Data Computation on Ceph Objects
Accelerating Data Computation on Ceph Objects
 
Alluxio Architecture and Performance
Alluxio Architecture and PerformanceAlluxio Architecture and Performance
Alluxio Architecture and Performance
 
Realtime Analytical Query Processing and Predictive Model Building on High Di...
Realtime Analytical Query Processing and Predictive Model Building on High Di...Realtime Analytical Query Processing and Predictive Model Building on High Di...
Realtime Analytical Query Processing and Predictive Model Building on High Di...
 
Hybrid Data Lake Architecture with Presto & Spark in the cloud accessing on-p...
Hybrid Data Lake Architecture with Presto & Spark in the cloud accessing on-p...Hybrid Data Lake Architecture with Presto & Spark in the cloud accessing on-p...
Hybrid Data Lake Architecture with Presto & Spark in the cloud accessing on-p...
 
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
RedisConf17- Zettaset + Redis - Protecting Redis Enterprise while Maintaining...
 
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
Intel: How to Use Alluxio to Accelerate BigData Analytics on the Cloud and Ne...
 
Data Orchestration for AI, Big Data, and Cloud
Data Orchestration for AI, Big Data, and CloudData Orchestration for AI, Big Data, and Cloud
Data Orchestration for AI, Big Data, and Cloud
 
Reducing large S3 API costs using Alluxio at Datasapiens
Reducing large S3 API costs using Alluxio at Datasapiens Reducing large S3 API costs using Alluxio at Datasapiens
Reducing large S3 API costs using Alluxio at Datasapiens
 
HPC DAY 2017 | FlyElephant Solutions for Data Science and HPC
HPC DAY 2017 | FlyElephant Solutions for Data Science and HPCHPC DAY 2017 | FlyElephant Solutions for Data Science and HPC
HPC DAY 2017 | FlyElephant Solutions for Data Science and HPC
 
Why Software-Defined Storage Matters
Why Software-Defined Storage MattersWhy Software-Defined Storage Matters
Why Software-Defined Storage Matters
 
HPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big DataHPC DAY 2017 | HPE Storage and Data Management for Big Data
HPC DAY 2017 | HPE Storage and Data Management for Big Data
 
Storage for big-data by Joshua Robinson
Storage for big-data by Joshua RobinsonStorage for big-data by Joshua Robinson
Storage for big-data by Joshua Robinson
 
Containerized Storage
Containerized StorageContainerized Storage
Containerized Storage
 
HPC DAY 2017 | The network part in accelerating Machine-Learning and Big-Data
HPC DAY 2017 | The network part in accelerating Machine-Learning and Big-DataHPC DAY 2017 | The network part in accelerating Machine-Learning and Big-Data
HPC DAY 2017 | The network part in accelerating Machine-Learning and Big-Data
 
cleversafe_definitive_guide_white_paper
cleversafe_definitive_guide_white_papercleversafe_definitive_guide_white_paper
cleversafe_definitive_guide_white_paper
 
The Architecture of Decoupling Compute and Storage with Alluxio
The Architecture of Decoupling Compute and Storage with AlluxioThe Architecture of Decoupling Compute and Storage with Alluxio
The Architecture of Decoupling Compute and Storage with Alluxio
 
Ibm power systems hpc cluster
Ibm power systems hpc cluster Ibm power systems hpc cluster
Ibm power systems hpc cluster
 
How Open Source Will Change How You Think about Storage - LGI Tech Summit
How Open Source Will Change How You Think about Storage - LGI Tech SummitHow Open Source Will Change How You Think about Storage - LGI Tech Summit
How Open Source Will Change How You Think about Storage - LGI Tech Summit
 
Red Hat Storage Day New York - Penguin Computing Spotlight: Delivering Open S...
Red Hat Storage Day New York - Penguin Computing Spotlight: Delivering Open S...Red Hat Storage Day New York - Penguin Computing Spotlight: Delivering Open S...
Red Hat Storage Day New York - Penguin Computing Spotlight: Delivering Open S...
 
Red Hat Storage Day Dallas - Why Software-defined Storage Matters
Red Hat Storage Day Dallas - Why Software-defined Storage MattersRed Hat Storage Day Dallas - Why Software-defined Storage Matters
Red Hat Storage Day Dallas - Why Software-defined Storage Matters
 

Similar to Webinar: Three Reasons Why NAS is No Good for AI and Machine Learning

Lessons learned processing 70 billion data points a day using the hybrid cloud
Lessons learned processing 70 billion data points a day using the hybrid cloudLessons learned processing 70 billion data points a day using the hybrid cloud
Lessons learned processing 70 billion data points a day using the hybrid cloud
DataWorks Summit
 
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
Hitachi Vantara
 

Similar to Webinar: Three Reasons Why NAS is No Good for AI and Machine Learning (20)

HPE Solutions for Challenges in AI and Big Data
HPE Solutions for Challenges in AI and Big DataHPE Solutions for Challenges in AI and Big Data
HPE Solutions for Challenges in AI and Big Data
 
Saviak lviv ai-2019-e-mail (1)
Saviak lviv ai-2019-e-mail (1)Saviak lviv ai-2019-e-mail (1)
Saviak lviv ai-2019-e-mail (1)
 
Lessons learned processing 70 billion data points a day using the hybrid cloud
Lessons learned processing 70 billion data points a day using the hybrid cloudLessons learned processing 70 billion data points a day using the hybrid cloud
Lessons learned processing 70 billion data points a day using the hybrid cloud
 
How the Development Bank of Singapore solves on-prem compute capacity challen...
How the Development Bank of Singapore solves on-prem compute capacity challen...How the Development Bank of Singapore solves on-prem compute capacity challen...
How the Development Bank of Singapore solves on-prem compute capacity challen...
 
Alluxio Monthly Webinar | Why NFS/NAS on Object Storage May Not Solve Your AI...
Alluxio Monthly Webinar | Why NFS/NAS on Object Storage May Not Solve Your AI...Alluxio Monthly Webinar | Why NFS/NAS on Object Storage May Not Solve Your AI...
Alluxio Monthly Webinar | Why NFS/NAS on Object Storage May Not Solve Your AI...
 
Webinar: The Four Requirements of a Cloud-Era File System
Webinar: The Four Requirements of a Cloud-Era File SystemWebinar: The Four Requirements of a Cloud-Era File System
Webinar: The Four Requirements of a Cloud-Era File System
 
NVMe and Flash – Make Your Storage Great Again!
NVMe and Flash – Make Your Storage Great Again!NVMe and Flash – Make Your Storage Great Again!
NVMe and Flash – Make Your Storage Great Again!
 
New Business Applications Powered by In-Memory Technology @MIT Forum for Supp...
New Business Applications Powered by In-Memory Technology @MIT Forum for Supp...New Business Applications Powered by In-Memory Technology @MIT Forum for Supp...
New Business Applications Powered by In-Memory Technology @MIT Forum for Supp...
 
IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015IBM Spectrum Scale Overview november 2015
IBM Spectrum Scale Overview november 2015
 
NGD Systems and Microsoft Keynote Presentation at IPDPS MPP in Vacouver
NGD Systems and Microsoft Keynote Presentation at IPDPS MPP in VacouverNGD Systems and Microsoft Keynote Presentation at IPDPS MPP in Vacouver
NGD Systems and Microsoft Keynote Presentation at IPDPS MPP in Vacouver
 
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
Hitachi solution-profile-advanced-project-version-management-in-schlumberger-...
 
S100299 ibm-cos-orlando-v1804c
S100299 ibm-cos-orlando-v1804cS100299 ibm-cos-orlando-v1804c
S100299 ibm-cos-orlando-v1804c
 
Accelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud EraAccelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud Era
 
Macroview Netapp Overview
Macroview Netapp OverviewMacroview Netapp Overview
Macroview Netapp Overview
 
NetApp All Flash storage
NetApp All Flash storageNetApp All Flash storage
NetApp All Flash storage
 
Accelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud EraAccelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud Era
 
Accelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud EraAccelerate Analytics and ML in the Hybrid Cloud Era
Accelerate Analytics and ML in the Hybrid Cloud Era
 
Webinar: The Bifurcation of the Flash Market
Webinar: The Bifurcation of the Flash MarketWebinar: The Bifurcation of the Flash Market
Webinar: The Bifurcation of the Flash Market
 
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based HardwareRed hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
 
Cloudian Webinar - 7 Key Reasons why Object Storage lowers Storage TCO
Cloudian Webinar - 7 Key Reasons why Object Storage lowers Storage TCOCloudian Webinar - 7 Key Reasons why Object Storage lowers Storage TCO
Cloudian Webinar - 7 Key Reasons why Object Storage lowers Storage TCO
 

More from Storage Switzerland

Webinar: Does Your Data Center Need NVMe?
Webinar: Does Your Data Center Need NVMe?Webinar: Does Your Data Center Need NVMe?
Webinar: Does Your Data Center Need NVMe?
Storage Switzerland
 

More from Storage Switzerland (20)

Webinar: Are You Treating Unstructured Data as a Second Class Citizen?
Webinar: Are You Treating Unstructured Data as a Second Class Citizen?Webinar: Are You Treating Unstructured Data as a Second Class Citizen?
Webinar: Are You Treating Unstructured Data as a Second Class Citizen?
 
Webinar: Five Reasons Modern Data Centers Need Tape
Webinar: Five Reasons Modern Data Centers Need TapeWebinar: Five Reasons Modern Data Centers Need Tape
Webinar: Five Reasons Modern Data Centers Need Tape
 
Special Presentation of Meet The CEOs - Commvault and Hedvig
Special Presentation of Meet The CEOs - Commvault and HedvigSpecial Presentation of Meet The CEOs - Commvault and Hedvig
Special Presentation of Meet The CEOs - Commvault and Hedvig
 
Panel Discussion: Is Computational Storage a Better Path to Extreme Performance?
Panel Discussion: Is Computational Storage a Better Path to Extreme Performance?Panel Discussion: Is Computational Storage a Better Path to Extreme Performance?
Panel Discussion: Is Computational Storage a Better Path to Extreme Performance?
 
Webinar: Complete Your Cloud Transformation - Store Your Data in The Cloud
Webinar: Complete Your Cloud Transformation - Store Your Data in The CloudWebinar: Complete Your Cloud Transformation - Store Your Data in The Cloud
Webinar: Complete Your Cloud Transformation - Store Your Data in The Cloud
 
Webinar: Simplifying the Enterprise Hybrid Cloud with Azure Stack HCI
Webinar: Simplifying the Enterprise Hybrid Cloud with Azure Stack HCIWebinar: Simplifying the Enterprise Hybrid Cloud with Azure Stack HCI
Webinar: Simplifying the Enterprise Hybrid Cloud with Azure Stack HCI
 
Webinar: Designing a Storage Consolidation Strategy for Today, the Future and...
Webinar: Designing a Storage Consolidation Strategy for Today, the Future and...Webinar: Designing a Storage Consolidation Strategy for Today, the Future and...
Webinar: Designing a Storage Consolidation Strategy for Today, the Future and...
 
Webinar: Is It Time to Upgrade Your Endpoint Data Strategy?
Webinar: Is It Time to Upgrade Your Endpoint Data Strategy?Webinar: Is It Time to Upgrade Your Endpoint Data Strategy?
Webinar: Is It Time to Upgrade Your Endpoint Data Strategy?
 
Webinar: Rearchitecting Storage for the Next Wave of Splunk Data Growth
Webinar: Rearchitecting Storage for the Next Wave of Splunk Data GrowthWebinar: Rearchitecting Storage for the Next Wave of Splunk Data Growth
Webinar: Rearchitecting Storage for the Next Wave of Splunk Data Growth
 
Webinar: Three Steps to Modernizing Backup Storage
Webinar: Three Steps to Modernizing Backup StorageWebinar: Three Steps to Modernizing Backup Storage
Webinar: Three Steps to Modernizing Backup Storage
 
Webinar: NAS vs Object - Can NAS Make a Comeback?
Webinar: NAS vs Object - Can NAS Make a Comeback?Webinar: NAS vs Object - Can NAS Make a Comeback?
Webinar: NAS vs Object - Can NAS Make a Comeback?
 
Webinar: NAS vs Object - Can NAS Make a Comeback?
Webinar: NAS vs Object - Can NAS Make a Comeback?Webinar: NAS vs Object - Can NAS Make a Comeback?
Webinar: NAS vs Object - Can NAS Make a Comeback?
 
Webinar: 5 Critical Enterprise Cloud Backup Capabilities
Webinar: 5 Critical Enterprise Cloud Backup CapabilitiesWebinar: 5 Critical Enterprise Cloud Backup Capabilities
Webinar: 5 Critical Enterprise Cloud Backup Capabilities
 
Webinar: Overcoming the Shortcomings of Legacy NAS with Microsoft Azure
Webinar: Overcoming the Shortcomings of Legacy NAS with Microsoft AzureWebinar: Overcoming the Shortcomings of Legacy NAS with Microsoft Azure
Webinar: Overcoming the Shortcomings of Legacy NAS with Microsoft Azure
 
Webinar: 3 Steps to be a Storage Superhero - How to Slash Storage Costs
Webinar: 3 Steps to be a Storage Superhero - How to Slash Storage CostsWebinar: 3 Steps to be a Storage Superhero - How to Slash Storage Costs
Webinar: 3 Steps to be a Storage Superhero - How to Slash Storage Costs
 
Webinar: Does Your Data Center Need NVMe?
Webinar: Does Your Data Center Need NVMe?Webinar: Does Your Data Center Need NVMe?
Webinar: Does Your Data Center Need NVMe?
 
Webinar: All in the Cloud - Data Protection Up, Costs Down
Webinar: All in the Cloud - Data Protection Up, Costs DownWebinar: All in the Cloud - Data Protection Up, Costs Down
Webinar: All in the Cloud - Data Protection Up, Costs Down
 
Webinar: How to Put an End to Hyperconverged Silos
Webinar: How to Put an End to Hyperconverged SilosWebinar: How to Put an End to Hyperconverged Silos
Webinar: How to Put an End to Hyperconverged Silos
 
15 Minute Friday: Tips for The Weekend - Stop the Unstructured Data Madness
15 Minute Friday: Tips for The Weekend - Stop the Unstructured Data Madness15 Minute Friday: Tips for The Weekend - Stop the Unstructured Data Madness
15 Minute Friday: Tips for The Weekend - Stop the Unstructured Data Madness
 
Webinar: 2019 Storage Strategies Series - What’s Your Plan for Object Storage?
Webinar: 2019 Storage Strategies Series - What’s Your Plan for Object Storage?Webinar: 2019 Storage Strategies Series - What’s Your Plan for Object Storage?
Webinar: 2019 Storage Strategies Series - What’s Your Plan for Object Storage?
 

Recently uploaded

Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Safe Software
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
Joaquim Jorge
 

Recently uploaded (20)

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
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024
 
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
 
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
 
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...
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemkeProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 
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
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Top 10 Most Downloaded Games on Play Store in 2024
Top 10 Most Downloaded Games on Play Store in 2024Top 10 Most Downloaded Games on Play Store in 2024
Top 10 Most Downloaded Games on Play Store in 2024
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
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
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
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...
 
Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 

Webinar: Three Reasons Why NAS is No Good for AI and Machine Learning

  • 1. Three Reasons Why NAS is No Good for AI and Machine Learning ● NAS can’t leverage today’s flash technology ● NAS has no or very rudimentary cloud integration ● NAS data protection schemes are expensive Watch and learn why: For audio playback and Q&A go to: bit.ly/NASAIML
  • 2. OurSpeakers Sean Kerr, Product Manager - HPC/AI Storage Solutions, HPE George Crump, Founder and Lead Analyst of Storage Switzerland Barbara Murphy, VP of Marketing, WekaIO
  • 3. What is Artificial Intelligence (AI) and Machine Learning (ML)?
  • 4. Top Mainstream Use Cases for AI and ML ● Autonomous Vehicles ● Fraud and Theft Detection ● Customer Service (ChatBots) ● Logistic Management ● Cyber-Security ● Smart Cities
  • 5. The Data Requirements for AI and ML ● Extreme High-Performance IO ● Low Latency ● Adept at High File Count ● Massive Scalability (Capacity and Performance) ● Multi-Location Processing (Multiple Data Centers and Cloud)
  • 6. AI and ML Tends to “Sneak Up” on Core IT ● Starts as a pilot or skunk-works project ● Uses existing storage resources or cloud storage ● Moves into production, breaking traditional storage ● IT tries to upgrade/expand traditional NAS ○ The project gets costly and can’t meet requirements
  • 7. 3 Reasons Why NAS Won’t Work for AI/ML
  • 8. NAS is a 90’S Technology ● Not architected to leverage flash ● Compute architecture has changed (GPUs + CPUs) ● NAS has to deliver high performance
  • 9. NAS Doesn’t Know Cloud ● Today’s NAS (file systems) have rudimentary cloud support ○ Typically replication for DR ● AI/ML Requires ○ Successful AI outcomes require keeping the compute layer saturated with data ● Seamless movement of data to the cloud for processing ○ Recall data (the results) from the cloud for processing ○ Archive data to the cloud for long-term storage
  • 10. NAS Data Protection is Expensive ● Impacts performance ○ Especially during failed-state ● Costly ○ Requires too much resource overhead (compute and capacity) ● Slow ○ Return to good-state requires either time or a lot of CPU
  • 11. ● Software not Hardware ● High Performance ○ Leverage Flash ○ Leverage NVMe IT Needs a Modern Approach ● More of a Fabric than a File System ○ Data moves seamlessly between locations and cloud ○ Exploits cloud tiers and compute ● Data Protection Designed for AI/ML
  • 12. 11© 2018 All rights reserved. How WekaIO Solves the AI Challenge Barbara Murphy VP of Marketing July 25, 2018 | 9:00 a.m. PDT
  • 13. 12© 2018 All rights reserved. NFS = Not For Speed GPU Bandwidth 11GBytes/sec/IB connection NFS Bandwidth 1 - 1.5GBytes/sec o A protocol invented in 1984 trying to solve a 2018 problem o pNFS tried to fix NFS but failed when metadata workloads exploded o Legacy parallel file systems like Lustre and GPFS cannot handle billions of small files – And they are complex to operate
  • 14. 13© 2018 All rights reserved. The Power of Parallelism o Modern networks on 100Gbit Ethernet are 100x faster than SSD o Everything can be distributed o With NVMe-oF, shared storage is faster than local storage
  • 15. 14© 2018 All rights reserved. WekaIO Solves the Data Accessibility Problem o Shared, POSIX compliant Parallel file system written for NVMe o Scales to trillions of files, billions in a single directory o Simple to install and use GPU Bandwidth 4-11GBytes/sec Per GPU Server 11GBytes/sec
  • 16. 15© 2018 All rights reserved. WekaIO Matrix: Full-featured and Flexible WekaIO Matrix Shared File System Fully Coherent POSIX File System That is Faster than Local File System Distributed Coding, More Resilient at Scale, Fast Rebuilds, End-to-End DP Instantaneous Snapshots, Clones, Tiering to S3, Partial File Rehydration InfiniBand or Ethernet, Converged or Dedicated Storage Server HPE AWS Cloud
  • 17. 16© 2018 All rights reserved. WekaIO is Faster than Local NVMe Drives Testing conducted on WekaIO/HPE reference platform • Inference benchmarks are I/O Bound • Single Client Performance on HPE Apollo 6500 • 8 NVIDIA V100 GPUs • WekaIO Matrix on 8x HPE DL360 • Local NVMe FS is I/O Bound http://dlpg.labs.hpe.com report 10 and 11
  • 18. 17© 2018 All rights reserved. Results From Deep Learning Use Case Actual measured data to the AI/ML GPU Cluster WekaIO is over 10x Faster than NFS All Flash NAS Initial tests Production
  • 19. HPE & WekaIO an Engineered Solution Sean Kerr
  • 20. WekaIO Matrix – Solution Building Block 19a00045311enw Apollo 2000 Gen10 ProLiant DL360 Gen10 Supported Storage Systems for WekaIO Matrix Apollo 6500 Gen10 Mellanox Switch WekaIO with Apollo 2000
  • 21. Thank you! Storage Switzerland http://www.storageswiss.com georgeacrump@storageswiss.com StorageSwiss on Twitter: http://twitter.com/storageswiss StorageSwiss on YouTube: http://www.youtube.com/user/storageswiss WekaIO https://www.weka.io info@weka.io WekaIO on Twitter: @Wekaio Barbara Murphy on Twitter: @scaleoutlady WekaIO on LinkedIN: https://www.linkedin.com/company/weka-io HPE https://www.hpe.com HPE on Twitter: https://twitter.com/hpe_hpc https://twitter.com/HPE_Servers HPE on Facebook: https://www.facebook.com/HPEServers/
  • 22. Three Reasons Why NAS is No Good for AI and Machine Learning For complete audio and Q&A please register for the On Demand Version: bit.ly/NASAIML