Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Technical Overview
Jeff Slapp
Director, Systems Engineering
Products:
• DataCore SANsymphony™-V
• DataCore Virtual SAN
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
HIGH-LEVEL OVERVIEW
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
 Widely deployed: Over 10,000 customers & Over 30,000 deployments
 Mature: 10th Generation & 18 years of development
3
Any physical host
DataCore + x86 = Enterprise Storage Controller
Any connection
Any hypervisor
Any application
Any storage
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 44
Storage Controller Architectures Compared
Expansion Slots*
CPUs** and Memory***
Controller 1
Expansion Slots
CPUs and Memory
Controller 1
Controller 2
Controller 2
Self-
Contained
Shared
Storage
Controllers
Expansion Slots*
CPUs** and Memory***
Expansion Slots
CPUs and Memory
Tightly-
Coupled
Operating
System
(Software)
Loosely-
Coupled
Operating
System
(Software)Controller Separation (> 100KM)
Discrete Non-
Shared
Storage
Controllers
No Controller Separation
Typical Storage Controllers
DataCore Storage Controllers
Storage Services Only (FC,
iSCSI, CIFS)
Storage (FC, iSCSI, CIFS) and
Application Services (Object, DB)
* Expansion slots only support specific devices
** CPU type is locked in and on vendor timetable
*** Memory is at a premium cost
Disk Is Single Point of Failure
No Single Point of Failure
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Traditional Converged Hyperconverged
Integrate, manage, and
enhance existing
storage
Leverage internal
storage, reduce
complexity and maintain
compute segregation
Consolidate all
functions for smallest
footprint and highest
performance
Same software, with an integrated management console across all three!
Deployment Model Independent
5
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Co-Existence of All Deployment Models
6
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
BREAKING WITH TRADITION
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Attacking The I/O Problem
• The traditional approach of dealing with the I/O problem is to push the I/O down to the disk.
• This means the only way to deal with increasing I/O demand is to add more disks and/or more
expensive disks (i.e. flash) to the architecture (Hardware Parallelization).
• The result of this is increased cost, size, and complexity, while not significantly impacting
response times.
• A better approach is handling the I/O as soon as it arrives at the system (I/O Parallelization).
Cost(Engines,Disks,Real
Estate,Environmentals,etc.)
Application I/O Demand
8
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
Understanding DataCore Parallel I/O
9
Parallel Application I/O
(Databases, Hypervisors)
Storage Admin: Performance is terrible, we
need to add more disks.
Serial Storage I/O
(Typical Storage)
Parallel Application I/O
(Databases, Hypervisors)
I/O Parallelization
Approach
Storage Admin: Performance is
unbelievable, takes up little space, and is
very affordable.
Parallel Application I/O
(Databases, Hypervisors)
Hardware Parallelization
Approach
Storage Admin: Latency is still very high,
takes up a lot of space, and this is getting
expensive.
Where Would You Rather Deal With Your I/O?
Latest Intel E5v4 Processors
194 GHz Parallel I/O Processing Power
across 88 Logical Processors with DDR4 RAM
Latest Intel E7v3 Processors
360 GHz Parallel I/O Processing Power
across 144 Logical Processors with DDR4 RAM
Closest To The Application With The Fastest Components?
LatencySeenByApplicationsandUsers
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 11
Results of Parallel I/O: Performance and Cost
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 12
Results of Parallel I/O : Real Estate
Hitachi VSP G1000 DataCore Parallel Server
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 13
Results of Parallel I/O : Latency
0.22
0.35
0.58
0.64
0.72
0.96
0.07141 0.06595 0.07612 0.08431 0.08749 0.09995
0
0.2
0.4
0.6
0.8
1
1.2
10% Load 50% Load 80% Load 90% Load 95% Load 100% Load
AverageResponseTime(ms)
SPC-1 Workload Generator
Ramp Phase Response Time / Throughput Curve
Hitachi VSP G1000 DataCore/Lenovo (SSD/HDD)
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
THE POSSIBILITIES
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 15
Enterprise Hybrid Services with Hyper-V
Copyright © 2016 DataCore Software Corp. – All Rights Reserved. 16
Enterprise Hybrid Services with VMware
THE BIG DATA EQUATION
HAS BEEN SOLVED
Physical Capacity
368 TB
Processing Capacity
194 GHz
=
+
Lenovo® x3650 M5 with Intel® Xeon® E5v4 Processors
Storage Performance
>1.5 Million IOps
Physical Capacity
7.73 PB
>31.5 Million
Combined
SPC-1 IOps
Native FC and iSCSI
Block Services
1,848 Logical
Processors
31.5 TBs of RAM
and High-Speed
Cache
Unified Compute
AND Storage
HDFS, Ceph, Lustre,
GlusterFS, xDFS,
NFS, CIFS
Big Data
Application
Agnostic
StorageCapabilitiesComputeCapabilitiesPlatformCapabilities
4,074 GHz of
Compute and
Storage I/O Power
1Rack(42U)
Copyright © 2016 DataCore Software Corp. – All Rights Reserved.
www.datacore.com
©2015 DataCore Software Corporation All rights reserved. DataCore, the
DataCore logo and SANsymphony are trademarks or registered trademarks of
DataCore Software Corporation. All other products, services and company names
mentioned herein may be trademarks of their respective owners.
THANK YOU

DataCore Technology Overview

  • 1.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. Copyright © 2016 DataCore Software Corp. – All Rights Reserved. Technical Overview Jeff Slapp Director, Systems Engineering Products: • DataCore SANsymphony™-V • DataCore Virtual SAN
  • 2.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. HIGH-LEVEL OVERVIEW
  • 3.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved.  Widely deployed: Over 10,000 customers & Over 30,000 deployments  Mature: 10th Generation & 18 years of development 3 Any physical host DataCore + x86 = Enterprise Storage Controller Any connection Any hypervisor Any application Any storage
  • 4.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 44 Storage Controller Architectures Compared Expansion Slots* CPUs** and Memory*** Controller 1 Expansion Slots CPUs and Memory Controller 1 Controller 2 Controller 2 Self- Contained Shared Storage Controllers Expansion Slots* CPUs** and Memory*** Expansion Slots CPUs and Memory Tightly- Coupled Operating System (Software) Loosely- Coupled Operating System (Software)Controller Separation (> 100KM) Discrete Non- Shared Storage Controllers No Controller Separation Typical Storage Controllers DataCore Storage Controllers Storage Services Only (FC, iSCSI, CIFS) Storage (FC, iSCSI, CIFS) and Application Services (Object, DB) * Expansion slots only support specific devices ** CPU type is locked in and on vendor timetable *** Memory is at a premium cost Disk Is Single Point of Failure No Single Point of Failure
  • 5.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. Traditional Converged Hyperconverged Integrate, manage, and enhance existing storage Leverage internal storage, reduce complexity and maintain compute segregation Consolidate all functions for smallest footprint and highest performance Same software, with an integrated management console across all three! Deployment Model Independent 5
  • 6.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. Co-Existence of All Deployment Models 6
  • 7.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. BREAKING WITH TRADITION
  • 8.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. Attacking The I/O Problem • The traditional approach of dealing with the I/O problem is to push the I/O down to the disk. • This means the only way to deal with increasing I/O demand is to add more disks and/or more expensive disks (i.e. flash) to the architecture (Hardware Parallelization). • The result of this is increased cost, size, and complexity, while not significantly impacting response times. • A better approach is handling the I/O as soon as it arrives at the system (I/O Parallelization). Cost(Engines,Disks,Real Estate,Environmentals,etc.) Application I/O Demand 8
  • 9.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. Understanding DataCore Parallel I/O 9 Parallel Application I/O (Databases, Hypervisors) Storage Admin: Performance is terrible, we need to add more disks. Serial Storage I/O (Typical Storage) Parallel Application I/O (Databases, Hypervisors) I/O Parallelization Approach Storage Admin: Performance is unbelievable, takes up little space, and is very affordable. Parallel Application I/O (Databases, Hypervisors) Hardware Parallelization Approach Storage Admin: Latency is still very high, takes up a lot of space, and this is getting expensive.
  • 10.
    Where Would YouRather Deal With Your I/O? Latest Intel E5v4 Processors 194 GHz Parallel I/O Processing Power across 88 Logical Processors with DDR4 RAM Latest Intel E7v3 Processors 360 GHz Parallel I/O Processing Power across 144 Logical Processors with DDR4 RAM Closest To The Application With The Fastest Components? LatencySeenByApplicationsandUsers
  • 11.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 11 Results of Parallel I/O: Performance and Cost
  • 12.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 12 Results of Parallel I/O : Real Estate Hitachi VSP G1000 DataCore Parallel Server
  • 13.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 13 Results of Parallel I/O : Latency 0.22 0.35 0.58 0.64 0.72 0.96 0.07141 0.06595 0.07612 0.08431 0.08749 0.09995 0 0.2 0.4 0.6 0.8 1 1.2 10% Load 50% Load 80% Load 90% Load 95% Load 100% Load AverageResponseTime(ms) SPC-1 Workload Generator Ramp Phase Response Time / Throughput Curve Hitachi VSP G1000 DataCore/Lenovo (SSD/HDD)
  • 14.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. THE POSSIBILITIES
  • 15.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 15 Enterprise Hybrid Services with Hyper-V
  • 16.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. 16 Enterprise Hybrid Services with VMware
  • 17.
    THE BIG DATAEQUATION HAS BEEN SOLVED Physical Capacity 368 TB Processing Capacity 194 GHz = + Lenovo® x3650 M5 with Intel® Xeon® E5v4 Processors Storage Performance >1.5 Million IOps Physical Capacity 7.73 PB >31.5 Million Combined SPC-1 IOps Native FC and iSCSI Block Services 1,848 Logical Processors 31.5 TBs of RAM and High-Speed Cache Unified Compute AND Storage HDFS, Ceph, Lustre, GlusterFS, xDFS, NFS, CIFS Big Data Application Agnostic StorageCapabilitiesComputeCapabilitiesPlatformCapabilities 4,074 GHz of Compute and Storage I/O Power 1Rack(42U)
  • 18.
    Copyright © 2016DataCore Software Corp. – All Rights Reserved. www.datacore.com ©2015 DataCore Software Corporation All rights reserved. DataCore, the DataCore logo and SANsymphony are trademarks or registered trademarks of DataCore Software Corporation. All other products, services and company names mentioned herein may be trademarks of their respective owners. THANK YOU