GLOBAL SPONSORS

High-end All Flah Storage

Milan ZofkaAdvisory Systems Engineer - MIT

And what’s up with NVMe?

XtremIOX2

Why XtremIO?

App Integrated CopiesRich Application Integration. No Compromise Copy services.

Unmatched EfficiencyMaximize Efficiency with Deduplication and Compression.

Consistent Performance Inline, all the time data services with no performance impact

Unique characteristics of XtremIO architectureSecret sauce in software

RAM

Scale-Out

Metadata

Content-Aware

Metadata

100% In-Memory

Inline Data Services

DATA

XtremIO X2 Inline Data Reduction In Action

XtremIO Data Processing

Input data stream

Inline deduplication

that is fast and scalable

ALWAYS-ON DEDUPLICATION AND COMPRESSION

Memory

Flash Media

22

Ingest Fingerprint

Analysis

2 323

4000Virtual Desktops hosted

per X-Brick

33%Lower $/desktop

25%Faster Boot Times

Up to

40%More VDI Users

80%Better application latency

2XBetter copy operations

25%Better data reduction

2X# of XtremIO Virtual Copies

4Xbetter rack density

Improved

Performance

TCO Savings

Per X-BrickImproved

Efficiency

1/3rd

Lower $/GB

Up to

Summary: Why XtremIO X2?

Multi-dimensional scaling

Active Controller

Active Controller

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

Scale-up X-Brick to

138TB raw

Active Controller

Active Controller

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

Active Controller

Active Controller

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

SSDFLASH

….........

Scale-out a cluster to 8 X-Bricks** of

1.1 PB raw

Up to

5.5PBof effective capacity*

*6:1 Data reduction

** >4 X-Bricks and adding X-Bricks in field supported post-GA

X-Brick configurations: X2-S & X2-R

Configuration X-Brick

Minimum Raw

X-Brick

Maximum Raw

X-Brick Typical

Max Effective*

Cluster Size

in X-Bricks

X2-S 7.2 TB 28.8 TB 96.8TB Up to 4

X2-R 34.5 TB 138.2 TB 464.3TB Up to 8**

*Assuming 4:1 DRR and 84% usable:raw

** Initial Max X2-R X-Bricks = 4. Post GA we will support 8

Compression improvements in X2X2 delivers ~25% improved compression compared to X1

Systems typically experiences:

~2-3:1*Compression ratio

Intelligent

Packing

Algorithm

*Max block compression ratio 16:1

PCIe NV-RAM UnitReplaces BBU in X1

• Used for data and metadata vaulting

• Increased reliability

• Reduced complexity

• Reduced cabling

• Reduce overall cluster RUs

• Leverages super capacitor

• Allows odd X-Brick supportSuper-Cap

App

Self

Service

App Integration &

orchestration

XtremIO Virtual Copies (XVC)

Consistent Multi-Dimensional Scale

Performance & Data Services

More Copies?

DEVELOPMENT PRODUCTION

ANALYTICS OPS+TEST

Why Not?

X2 Solves the CDM ProblemiCDM - INSTANT REFRESH, ANY DIRECTION

XtremIO Metadata-aware native replication

• Uses XtremIO in-memory snapshots

• Wizard based

• Full operational disaster recovery

• RPO as low as 30 seconds

• Immediate RTO

• Up to 1000 recover points

• “Fan-in” configurations

• Supports XtremIO High Performance

• Efficient Metadata-aware Replication

• Efficient replication - Compression

aware

Easy Operation Best Protection Superior Performance

XtremIO native replication

75%Data reduction

RPOsas low as

30seconds

=S1S2

DELTA(S1, S2)

222

Primary Site DR Site

Deduplicated and net new

blocks transferred

Just pointer

updates for

existing blocks

With up to…

Redesigned user interface

• Simple and intuitive

• 1-2-3 Provisioning

• Tagging and Search

• X1 and X2 support Multiple clusters

• HTML5 (no Java) Nothing to install

Popular browser support

VMAX ALL FLASH

Dell EMC VMAX All Flash

• Modern consumption model,

simple packaging and pricing

• Engineered for all-flash with

high capacity enterprise SSD’s

• Optimized O/S for superior

response time (.5ms)

• Designed for mission-critical

workloads for every business

• Inline H/W Compression with

all data services & encryption

WHAT’S IMPORTANT

• High availability, on-line

upgrades, and data integrity

• Trusted data services for

mission-critical apps @ scale

• Delivered as standalone or

converged infrastructure

• Extensive ecosystem support,

qualifications, & validation

• Dell EMC premium support

with 24x7 service coverage

WHAT’S PROVEN

The VMAX All Flash Family

Software Package Highlights

SnapVX

Compression

F SOFTWARE

FX SOFTWARE Above +

ViPR Suite

PowerPath/VE

and more…

SRDF

D@RE

eNAS

1M IOPSRRH-8K

1PBe Capacity

64 FC/iSCSI Ports

1 to 2 V-Bricks

6.7M IOPSRRH-8K

4PBe Capacity

192 FC/iSCSI or 256 FICON Ports

1 to 8 V-Bricks

VMAX 950FVMAX 250F

Optimized for flash

IOPS and throughput

Massive increase in

per-port performance

Balanced performance

across CPU resources

Front-end (FA) ports

Infrastructure

Management (IM)

Back-end (DA) ports

Front-end Core Pool

Back-end Core Pool

4

0 1 N-1 N

Enginuity Data

Services (EDS)

24

Architected for all flashV-Bricks optimized for multi-core CPUs

5 6 7 8 26 2725

VMAX All-Flash Engines

250F 950F

Broadwell 12 core CPU 2.2GHz

48 CPU cores per engine

Up to 2 TB cache per engine

Up to 32 FE ports per engine

Broadwell 18 core CPU 2.3GHz

72 CPU cores per engine

Up to 2 TB cache per engine

Up to 24 FE ports per engine

Introducing VMAX 950F

• VMAX 950F/FX

– 6.7M IOPS**. 350µs response time for OLTP**

– 4PB effective capacity

– Up to 576 CPU Cores (4 x 18-core CPUs per V-Brick)

• Supports 1 to 8 V-Brick

– 53TB usable V-Brick (Includes 1 engine & 53TBu)

– 13TB usable Flash Packs (add-on)

• Up to192 front end host port open / 256 Mainframe

• Inline compression (open only) with all data services

• Mixed open and mainframe workloads

• Simple, appliance like software packaging – F & FX

Faster than

competition**

4X

Smaller footprint*(Performance & Capacity)

~25%

Better response time*

~30%

* Compared to VMAX 850F

** Based on internal testing vs. Hitachi (IOPS for OLTP w/ compression)

VMAX All Flash 250F/FX Configuration Details

• V-Bricks in single increments

– Redundant dual director engine design

– 48 Broadwell CPU cores @ 2.2GHz

• Up to 4 I/O module pairs per V-Brick

– Each 4x 16Gb FC or 10Gb iSCSI› NO mainframe (FICON) support

– eNAS 10Gb IP

• 2 DAEs per V-Brick (12Gb SAS)

– 50 x 2.5” flash drives per V-Brick

• RAID 5 (3+1, 7+1) or RAID 6 (6+2)

VMAX Customer Read Hit Statistics

Read hit percentage

Dis

trib

utio

n a

cro

ss s

yste

ms

Average 54%

Cache hits reduce

latency to 0.1-0.3ms

100% of writes are cached

~50% of reads are cached

Daily average read hits

from 2,000+ VMAX systems

Reads misses serviced by

flash storage (<1ms)

VMAX All-Flash Online Code Updates: True NDUUnique in the industry

• < 10 second array OS upgrade

• No component downtime

– No rolling outage upgrade

– No Failover/Failback processes involved

– No switching LUN ownership/trespass required

• Ports never drop light

– Servers never see logout/login (no fabric RSCN)

• Online downgrades work the same way

• Historical feature going back

many generations

DOWNTIME COST

$1.8 millionper day

$45,000 per hour

$750 per minute

“Thank you VMAX for giving me back my weekends”

“VMAX NDU is the gold standard for upgrades”

“Nobody knows its happening – it just works”

Remote Replication Gold Standard – SRDF

Synchronous AsynchronousMetro

Zero data loss

High performance

Scalable consistency

Extended distance

Multi-cycle mode

Remote link resiliency

Active/active

Automated failover/back

Non-disruptive migrations

2-site, 3-site, and 4-site replication

Simple: <2 minutes to configure

Up to 100km

1

42

3

Unlimited Distance

1

23

Witness

Up to 100km

VMAX All-Flash SnapVXLocal replication designed for performance at scale

Awesome performance

Minimal additional capacity

Create 1000s of copies

For backup, test, dev, training, …

Integrated with apps/DBs

Easy, consistent PiT copies 0

5 000

10 000

15 000

20 000

25 000

30 000

35 000

40 000

45 000

50 000

Read IOPS Write IOPS Aggregate IOPS

OLTP workload before and After snapshots creation

OLTP Baseline OLTP Workload and Snapshot Creation

Only ~3% impact

Under heavy load

Secure Snaps with TimeFinder SnapVXSnaps that stay

• All snaps created with retention period

– Automatic delete on expiration

• Secure snaps are protected

– Expiration cannot be shortened

• Protection against changes

– Accidental or malicious

– Helps internal governance

1 day retention

3 day retention

7 day retention

Accident

ProtectPoint Storage-integrated ProtectionDramatically faster backup and recovery

Faster

backup & recovery

Eliminate

application impact

Reduce

cost and complexity

20x Faster

Backup

10x Faster

Recovery

RecoverPoint for VMAX All Flash

• Snap-based replication

– No performance impact

– Adjust snap frequency per RPO

• Heterogeneous support

– VMAX to Unity, XtremIO, VPLEX

• Leverages RecoverPoint capabilities

– Concurrent local and remote replication

– Very low WAN bandwidth consumption

– Well orchestrated DR workflows

– Multi site configurationsProduction site

Local

Copy

Production

Remote site

Remote

Copy

VMAX and XtremIO All-Flash Storage EfficiencyReduces TCO

I n l i n e c o m p r e s s i o n

S n a p s

T h i n p r o v i s i o n i n g

Z e r o s p a c e r e c l a i m

4:1S t o r a g e E f f i c i e n c yC o m p r e s s i o n

2:1*

* Compression rates vary depending on customer applications and environments.

2:1 compression ratio is expected for typical OLTP workloads.

B l o c k

F i l e

Storage and NVMe

Emergent Non-volatile Media ImpactAddresses memory/storage latency/capacity gaps

10ns 1us 10us1ns 100ns 100us 1ms 10ms

HDD

Medium NV Media(DRAM)

(Processor

SRAM)

Memory access semantics

IO block access semantics

Low Speed Storage

Access

Latency

Rela

tive c

apacity

(not to

scale

)

Capacity/Latency Gap Fill

Faster

NV Media

$$$$ $$$ $$ $ < $ << $

Emergent Memory Domain

MLCSLC

TLCQLC

High Speed Storage

(NAND Flash)

Slower NV Media

NVM Express and I/O Latency

Source: Storage Technologies Group, Intel. Comparisons between memory technologies based on in-market product specifications and internal Intel specifications.

Latency

HDD+ SAS

NAND+ SAS

NAND+ NVMe

Drive

Latency

Controller

Latency(i.e. SAS HBA)

Software LatencySCM

+ NVMe

NVMe drives down connection latency

Storage-Class Memory technology offers ~10xlatency reduction versus NAND

NAND technology offers ~25x latency

reduction versus HDD

Slide credit: Intel and NVM Express

VMAX Storage Technology Evolution

* In Planning, final delivery dates are TBD

Dell EMC is working closely with NVMe and Storage-Class Memory suppliers and will be a leader in integrating, optimizing, and delivering next generation flash solutions

SCSI + HDD

1988

Industry’s first Intelligent Cached Disk Array combining cache and commodity HDDs

SAS + SLC

2008

Industry’s first Enterprise Array to

support SSD Flash and automated tiering

NVMe + NAND

Next*

Leadership for Enterprise Array delivering NVMe-connected SSDs

NVMe + SCM

Future*

Leadership for Enterprise Array delivering NVMe-connected SCM

~15->1 ~1.2->1 ~5->1

Thank you