Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
WEBINAR
NVMe and Flash –
Make Your Storage Great Again!
Alfons Michels
Senior Product Marketing Manager
Alex Best
Director, Technical Business Development
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 2
Data and Performance Requirements grow
much faster than Budgets
2012 2017 2022
Your data nearly doubles every 2 years
Your storage budget growths around
5% per year
Mind the
gap!
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
• What are my pain-points?
• Performance
• Rackspace / Operational cost / Energy
• Where does Flash help me?
• 10% of Data generate 90% of I/Os
• A Flash TB is still x-times more expensive than an HDD TB
• What is the performance guarantee?
• What is a sustainable / expandable entry into Flash?
3
What should You Consider when You Plan to
change to Flash/SSD Storage?
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 4
• How to integrate the
new technology into my
existing environment?
• Which new processes
are necessary to
manage my new
investments?
• Does each application
needs its dedicated
flash-storage?
• How to ensure essential
operational capabilities
like high availability or
transparent failover?
• What tools are available
in regard to data
migration, tiering and
thin provisioning?
Integration and
Management
Typical Challenges with Flash-Integration
Availability of
Functionality
Shared usage of
Flash Storage
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 5
• Dedicated insertion of
Flash
• Cost Effective
• Measurable
• Seems to be simple to
integrate
• Does generate (many)
Islands to Manage
• Cost Effective
• Performance
Improvement
• Seems to be difficult to
be Realized
• Does it really help My
most Demanding Apps?
• Seems to ensure best
Performance Results
• Very cost intensive
• No exactly predictable
Performance Advantage
• Entire Migration Needed
The most common ways
to add Flash to Your Application(s)
Isolate Applications
– use Local Flash
Rip and Replace
towards All Flash
Use Hybrid (tiered)
Storage (x% Flash)
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 6
• Shared usage of Flash
between server and
application
• Minimized downtimes
and risk at integration
• Simple expansion of
flash at any time
• Real path to all flash
integration options
• Complete set of all
storage services also for
flash
• Auto-Tiering on block-
level
Simple Integration and Shared Usage of
Flash by leveraging DataCore SDS
Hyperconverged
Environments
Performance
Acceleration Layer
Dedicated Flash
Storage
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 8
Flash is Fast, but what about its Connection?
-
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
Flash in traditional hybrid-array
Min. Latency Avg. Latency Max. Latency
Min. Read Latency Avg. Read Latency Max. Read Latency
Min. Write Latency Avg. Write Latency Max. Write Latency
-
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
Flash using DataCore SDS transport
Min. Latency Avg. Latency Max. Latency
Min. Read Latency Avg. Read Latency Max. Read Latency
Min. Write Latency Avg. Write Latency Max. Write Latency
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
9
Serial Scheduling DataCore Parallel I/O
Concurrent
tasks
End-to-End Parallelism with Parallel I/O
NVMe takes advantage of parallel, low latency data paths to underlying media, Parallel I/O
takes advantage of multi-cores to remove serial I/O bottlenecks prior to controllers/data paths
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 10
Data-Caching and -Tiering
Less active data will be
stored on cost-effective storage
Active data will be
stored on faster storage
Flash/
SAS
SAS/
SATA
Flash/
NVMe Cloud
Hot ColdWarm Archive
RAM-Cache
Burst
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 11
Auto-Tiering on Block-level
• Not all Tier-1-Applications have data
which needs the most performant
storage
• DataCore predictable ensures that only
the most challenging data are stored in
Tier-1-Storage
Application Data
Often
Accessed
Normally
Accessed
Rarely
Accessed
Tier-1 Tier-2 Tier-3 Tier-4 / Archive
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
Live Demo
12
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 13
Live Demo Setup
V-Cluster-2V-Cluster-1 Centos7-1 Centos7swarm
Flash
Storage
DataCore
Server Group
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
Real Life Example
14
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 16
• Provider of Cloud Services &
Software for Accounting
• Approx. 1000 employees, own datacenter
• Business Requirements
• Overcome Performance Complains at
peak times, e.g. with VAT declarations
• Plan to grow Cloud Business 4 Times
• Budget Constraints
• Initial Situation (storage)
• 2 x 90TB HDD (Raid 5 each, mirrored) +
2% SSDs as cache
Real Life Example
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 17
The Solution
Application Layer
Cloud Services
(unchanged)
Tier 1: 2x DataCore SDS
2x SN260 NVMe (12,8TB)
Tier 2: Classical Flash
2x 2U24 All Flash (SAS SSD,
Raid 5, 184 TB)
Tier 3: Classical HDD
2x Ultrastar Data60 (HDD,
Raid 5, 360 TB)
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
Compared to previous storage setup, this customer achieved:
• 600% more capacity
• >10 times better Performance
• 10TB rebuild with less than 1hr at bandwidth greater 1.5GB/s
• Measured using GrafANA test platform
• 30-40% lower TCO (over 5 years)
• Compared to continuing (w/o extension!) with the old infrastructure
18
Customer Benefits Achieved
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
Summary and Next Steps
19
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 20
• All-Flash
• 80 TB Flash
With an assumed HW-Price difference (NVMe Flash vs. SAS HDD)
of $900 per TB, You could save (64 x $900) $57,600!
• Hybrid (20% flash)
• 16 TB NVMe Flash
• 64 TB SAS HDD
Pointed usage of NVMe Flash:
Full Performance, tremendous HW Savings
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 21
Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 22
Various Deployment Options
Traditional
Converged
(Server-SAN)
Hyperconverged Hybrid-Converged
With seamless migration between them during regular business operation!
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
• Schedule a Meeting with our Solution Architects (just click
the Link) to figure out, how to integrate Flash via NVMe:
• Low Initial Investment and high investment protection . . .
• Low Total Cost of Ownership (TCO) and simple management . . .
• More Flexibility with Flash and NVMe simply be integrated. . .
• Better Economic Efficiency. . .
• High Performance for business-critical applications. . .
23
Next Step
Copyright © 2018 DataCore Software Corp. – All Rights Reserved.
Questions?
info@datacore.com
www.datacore.com

More Related Content

NVMe and Flash – Make Your Storage Great Again!

  • 1. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. WEBINAR NVMe and Flash – Make Your Storage Great Again! Alfons Michels Senior Product Marketing Manager Alex Best Director, Technical Business Development
  • 2. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 2 Data and Performance Requirements grow much faster than Budgets 2012 2017 2022 Your data nearly doubles every 2 years Your storage budget growths around 5% per year Mind the gap!
  • 3. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. • What are my pain-points? • Performance • Rackspace / Operational cost / Energy • Where does Flash help me? • 10% of Data generate 90% of I/Os • A Flash TB is still x-times more expensive than an HDD TB • What is the performance guarantee? • What is a sustainable / expandable entry into Flash? 3 What should You Consider when You Plan to change to Flash/SSD Storage?
  • 4. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 4 • How to integrate the new technology into my existing environment? • Which new processes are necessary to manage my new investments? • Does each application needs its dedicated flash-storage? • How to ensure essential operational capabilities like high availability or transparent failover? • What tools are available in regard to data migration, tiering and thin provisioning? Integration and Management Typical Challenges with Flash-Integration Availability of Functionality Shared usage of Flash Storage
  • 5. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 5 • Dedicated insertion of Flash • Cost Effective • Measurable • Seems to be simple to integrate • Does generate (many) Islands to Manage • Cost Effective • Performance Improvement • Seems to be difficult to be Realized • Does it really help My most Demanding Apps? • Seems to ensure best Performance Results • Very cost intensive • No exactly predictable Performance Advantage • Entire Migration Needed The most common ways to add Flash to Your Application(s) Isolate Applications – use Local Flash Rip and Replace towards All Flash Use Hybrid (tiered) Storage (x% Flash)
  • 6. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 6 • Shared usage of Flash between server and application • Minimized downtimes and risk at integration • Simple expansion of flash at any time • Real path to all flash integration options • Complete set of all storage services also for flash • Auto-Tiering on block- level Simple Integration and Shared Usage of Flash by leveraging DataCore SDS Hyperconverged Environments Performance Acceleration Layer Dedicated Flash Storage
  • 7. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 8 Flash is Fast, but what about its Connection? - 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 Flash in traditional hybrid-array Min. Latency Avg. Latency Max. Latency Min. Read Latency Avg. Read Latency Max. Read Latency Min. Write Latency Avg. Write Latency Max. Write Latency - 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 Flash using DataCore SDS transport Min. Latency Avg. Latency Max. Latency Min. Read Latency Avg. Read Latency Max. Read Latency Min. Write Latency Avg. Write Latency Max. Write Latency
  • 8. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 9 Serial Scheduling DataCore Parallel I/O Concurrent tasks End-to-End Parallelism with Parallel I/O NVMe takes advantage of parallel, low latency data paths to underlying media, Parallel I/O takes advantage of multi-cores to remove serial I/O bottlenecks prior to controllers/data paths
  • 9. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 10 Data-Caching and -Tiering Less active data will be stored on cost-effective storage Active data will be stored on faster storage Flash/ SAS SAS/ SATA Flash/ NVMe Cloud Hot ColdWarm Archive RAM-Cache Burst
  • 10. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 11 Auto-Tiering on Block-level • Not all Tier-1-Applications have data which needs the most performant storage • DataCore predictable ensures that only the most challenging data are stored in Tier-1-Storage Application Data Often Accessed Normally Accessed Rarely Accessed Tier-1 Tier-2 Tier-3 Tier-4 / Archive
  • 11. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. Live Demo 12
  • 12. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 13 Live Demo Setup V-Cluster-2V-Cluster-1 Centos7-1 Centos7swarm Flash Storage DataCore Server Group
  • 13. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. Real Life Example 14
  • 14. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 16 • Provider of Cloud Services & Software for Accounting • Approx. 1000 employees, own datacenter • Business Requirements • Overcome Performance Complains at peak times, e.g. with VAT declarations • Plan to grow Cloud Business 4 Times • Budget Constraints • Initial Situation (storage) • 2 x 90TB HDD (Raid 5 each, mirrored) + 2% SSDs as cache Real Life Example
  • 15. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 17 The Solution Application Layer Cloud Services (unchanged) Tier 1: 2x DataCore SDS 2x SN260 NVMe (12,8TB) Tier 2: Classical Flash 2x 2U24 All Flash (SAS SSD, Raid 5, 184 TB) Tier 3: Classical HDD 2x Ultrastar Data60 (HDD, Raid 5, 360 TB)
  • 16. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. Compared to previous storage setup, this customer achieved: • 600% more capacity • >10 times better Performance • 10TB rebuild with less than 1hr at bandwidth greater 1.5GB/s • Measured using GrafANA test platform • 30-40% lower TCO (over 5 years) • Compared to continuing (w/o extension!) with the old infrastructure 18 Customer Benefits Achieved
  • 17. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. Summary and Next Steps 19
  • 18. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 20 • All-Flash • 80 TB Flash With an assumed HW-Price difference (NVMe Flash vs. SAS HDD) of $900 per TB, You could save (64 x $900) $57,600! • Hybrid (20% flash) • 16 TB NVMe Flash • 64 TB SAS HDD Pointed usage of NVMe Flash: Full Performance, tremendous HW Savings
  • 19. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 21
  • 20. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. 22 Various Deployment Options Traditional Converged (Server-SAN) Hyperconverged Hybrid-Converged With seamless migration between them during regular business operation!
  • 21. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. • Schedule a Meeting with our Solution Architects (just click the Link) to figure out, how to integrate Flash via NVMe: • Low Initial Investment and high investment protection . . . • Low Total Cost of Ownership (TCO) and simple management . . . • More Flexibility with Flash and NVMe simply be integrated. . . • Better Economic Efficiency. . . • High Performance for business-critical applications. . . 23 Next Step
  • 22. Copyright © 2018 DataCore Software Corp. – All Rights Reserved. Questions? info@datacore.com www.datacore.com