EMC Vmax3 tech-deck deep dive

1EMC CONFIDENETIAL – INTERNAL AND PARTNER USE ONLY
TITLE ALL CAPS
INTERNAL, DARK BACKGROUND
Ensuring the Technical Win
Technical Marketing

StudentGuide&Internal&ConfidentialUpdateDaily
https://goo.gl/VVmVZ0
• Download this slide
• http://ouo.io/FiEuqy
EMC VMAX Technical Overview http://ouo.io/lBomA
Presentation - Symmetrix VMAX Family with Enginuity
5876 http://ouo.io/qkCML
Presentation - FTS Deep Dive
http://ouo.io/hhqO7
c
Presentation - VMAX Hardware Deep Dive http://ouo.io/GlO0v
EMC VMAX Technical Deep Dive http://ouo.io/FiEuqy
VMAX MF Information Experience_Feb23 http://ouo.io/Zz1ht

VMAX3 Architecture
Remote Replication
Local Replication
Service Level Objectives
Embedded File
Management
ProtectPoint
This Technical Deck is used to supplement the VMAX Sales 101 and 201 Decks

5EMC CONFIDENETIAL – INTERNAL AND PARTNER USE ONLY 5© Copyright 2014 EMC Corporation. All rights reserved.© Copyright 2014 EMC Corporation. All rights reserved.
REVOLUTIONARY NEW CONVERSATIONS
NEW
ARCHITECTURE
REVOLUTIONARY
OS
VMAX3
SYMMETRIX
MAINFRAME
DMX
CLIENT
SERVER
VMAX
VIRTUAL
ENVIRONMENTS
NEW PRICING
AND PACKAGING
PRICING
NEW USE
CASES
CAPABILITIES
HYBRID CLOUD

• Add value to the business
– Time to deploy new applications –
Ease of provisioning
• Reduce Cost
– Cloud – Scale to 100’s of thousands of
objects, multi box SRDF protection for
the cloud scale deployments
– Market leading performance (latency,
IOPs and Bandwidth)
– Agility- SLOs, ease of use
– Deliver consistent service levels –
SLOs maintained over time and
through tech refresh with future VMAX
technology
CORE BUSINESS VALUE PROPS
TheWorldsMostResilientStorageSystemisNowtheSimplest
• Reduced Complexity
– Ease of provisioning
(SLO), management
(Unisphere),
configuration
• Manage Risk
– Always on availability –
hardware architecture,
True NDU, etc.
– Consistent service level
delivery, always, with
true SLOs
VMAX Value

• SLO’s – Radical simplicity,
less time tuning, and ease of
management with Unisphere
• Snap scale / Pre-Dupe (more
concurrent snaps and less
cache usage)
• Revolutionary backup
(ProtectPoint)
• Performance and scale –
dynamic Virtual Matrix,
HYPERMAX
CORE TECH VALUE PROPS
VMAX3 Differentiation compared to Legacy VMAX
VMAX Value
• Density – lot more disks,
smaller batteries,
compute and disk density
(120 drives in 3U)
• Standard 3rd party rack
support
• Local Raid – flexible floor
planning, performance

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint
VMAX3 Architecture
8© Copyright 2014 EMC Corporation. All rights reserved.© Copyright 2014 EMC Corporation. All rights reserved.

Minimize risk, at cloud scale
• Able to put your eggs in one
basket: Always On
– 6x9’s (with SRDF)
– Redundant hardware
– Resiliency / Recoverability
(e.g.. T10DIF+ protects
against silent corruptions) –
VMAX is the most deployed
system for mission critical
apps
– Full NDU, with no loss of light
– Supplement with SRDF or
VPLEX as needed
CORE TECH VALUE PROPS- ARCHITECTURE
Minimize Risk and Reduce Cost and Complexity
Reduce Cost and Complexity
• Stop appliance sprawl with
Embedded apps
– Greater ease of use
– Less hardware to go wrong

CONSOLIDATE AT SCALE
100K 400K200K
2to4Controllers
48 Cores
2TB Cache
1,4402.5”Drives
64 Ports
500TB
2 to 8 Controllers
128 Cores
8TB Cache
2,880 2.5” Drives
128 Ports
2PB
2 to 16 Controllers
384 Cores
16TB Cache
5,760 2.5” Drives
256 Ports
4PB

94%OF CUSTOMERS OVER 5
YEARS SERVICE HAVE
100% RUN TIME
90%
OF UNITS RUNNING AT
100% AVAILABILITY
VMAX is
The Storage
Of Choice For Most
Enterprises
94% OF THE FORTUNE 50
RUN EMC VMAX

• Robust cache protection
– Mirrored Writes; ongoing cache integrity checks
• Advanced error detection
– T10DIF+; protects from data and control faults
• Preemptive event driver
– Prioritized processes; optimize error handling and host IO
over internal processes for improved performance and
reliability
• Redundancy at every level in the system
• True Non-Disruptive Hardware and Software
Upgrades
WHY VMAX3 – MISSION CRITICAL AVAILABILITY
Turn on when you deploy it! Turn off when you retire it!

• < 10 seconds to upgrade the OS across
the entire array – FAST IML
• Absolutely no component downtime for
an HYPERMAX NDU
• No rolling outage upgrade
• True active active; No Failover – Failback
processes involved
– No switching LUN ownership (trespass)
required
VMAX3 True NDU
Unique in the Industry
DOWNTIME COST
$1.8 million
per day
$45,000
per hour
$750
per minute
* Enterprise Management Associates - 2012

• All VMAX3 arrays are 100% virtually provisioned
• All VMAX3 arrays are shipped pre-configured
– TDATs created
– Virtual Pools created
– Default Storage Resource Pool created
– Service Level Objectives prepackaged
• FAST
– is always on, even in simple configurations
• Manages internal load balancing
• Includes TimeFinder snapshot data and SRDF/A Delta Set
Extension data
IMPORTANT TO KNOW

• Differentiation by type & quantity of Engines
– i.e. Performance, System Drive Count, Capacity
• Rack configurations with one or two engines
• Single engine systems have no fabric
– Multi-engine Systems will have Fabric
– Online Fabric Upgrade when Engine # 2 is added
• System Bay Dispersion
– Up to 25 Meters from System Bay 1
COMMON PLATFORM FEATURES (1)

• Vault to Flash in the Engine
– Flash IO module
– No vault drives
– No battery backup on any drives
• Up to 4 FE IO modules per director
– Up to 256 ports in a single 8 engine array
• 4-Port, 8Gbps or 16 Gbps FC IO modules
• 1GbE and 10GbE Modules (for use with
eNAS and SRDF)
• Support for Third Party Racking

• System Configuration Rules are the same
– Max 720 2.5” drive / 360 3.5” drives per
engine
– IO module population order
– Drive / RAID Protection offerings
• RAID1, RAID5 3+1 and 7+1, RAID6 6+2 and 14+2
– Dispersion capability
• Mixing 2.5” & 3.5” DAEs behind an Engine

• Available on 100K, 200K, & 400K
• Full complement of VMAX data
services
• Fully scalable (same maximum
Engine and Drive count limits as
hybrid configurations)
• Upgradeable/Convertible to hybrid in
the future
VMAX3 ALL FLASH CONFIGURATIONS

KEY VMAX3 SCALABILITY PARAMETERS
Item Limit
Track Size 128K
Max Device Size 16TB*
Host Addressable Devices (TDEVs) 64K*
Max Storage Groups 4000
Max Snaps per Source / Mounted
(Linked Targets)
256 / 1024
Host I/O Limits (Max per SG) 4K
SRDF Groups per Director / Array 250 / 250
* Will be increased post-GA

• The following software upgrades are Supported
– Addition of new array software (TF, SRDF and/or
ProtectPoint)
– Conversion of Suites: to either the Foundation Suite or the
Advanced Suite from a lower Suite (this may be deferred)
• The following hardware upgrades are supported:
– Addition of I/O Modules into empty slots
– Drive upgrades in empty slots (2015)
VMAX UPGRADE OPTIONS

Unisphere:
Unisphere for VMAX
Unisphere for File
Workload Planner
DB Storage Analyzer
Unisphere Suite
Order Default
HYPERMAX OS:
Enginuity, Migration
Tools, Auto provisioning
and CLI mgmt
Open Bid
Low Cost RFI
CLI mgmt only
HYPERMAX OS:
and CLI mgmt
If Multi-Tier Disks configured If Single -Tier Disks configured
Base Suite
Open Bid
Low Cost RFI
CLI mgmt only
HYPERMAX OS:
and CLI mgmt
FAST:
Data Placement
Optimization
RDF FAST Awareness
GUI mgmt
HYPERMAX OS:
and CLI mgmt
Foundation
Suite
Unisphere:
Unisphere for VMAX
Unisphere for File
Workload Planner
DB Storage Analyzer
FAST:
Data Placement
Optimization
RDF FAST Awareness
Advanced FAST:
SLOs, Array Advisor
HYPERMAX OS:
and CLI mgmt
FAST:
Data Placement
Optimization
RDF FAST Awareness
Advanced Suite
Order Default
Unisphere:
Unisphere for VMAX
Unisphere for File
Workload Planner
DB Storage Analyzer
SOFTWARE PACKAGING

Optional eNAS software (basic NAS protocols)
Total Productivity Pack (Q4)
Local + Remote Replication + VMAX Advanced Suites
Management
Advanced Suite
Foundation + Advanced FAST with SLO Provisioning
Foundation Suite
Base + Unisphere (VMAX + File)
Base Suite
FAST Base (single policy)
Remote Replication Suite (Q4)
SRDF/S, SRDF/A, SRDF/CE, File Replication
Protection
Local Replication Suite (Q4)
TimeFinder SnapVX, SnapSure
ProtectPoint (Q4)
Direct Backup/Recovery with DataDomain
HYPERMAX OS, Solutions Enabler, Auto Provisioning, Migration Tools
VMAX3 SOFTWARE CHOICES

Director 2
Director 1
• 4U Enclosure
• 2 Directors
• Each Director has
Redundant Power &
Cooling
• 2 Management Modules*
• 11 IO Slots per Director
• 2 or 4 Vault to Flash
SLICs in each director
– Compulsory item
NEW ENGINES
Director 2
Director 1
*Management modules only required in first engine

NEW ENGINES
100K 200K 400K
Processor /
Director
Dual Processor
Ivy Bridge
Dual Processor
Ivy Bridge
Dual Processor
Ivy Bridge
Cores/Engine 24 32 48
Frequency 2.1GHz 2.6GHz 2.7GHz
Memory/
engine
512GB, 1TB
512GB, 1TB,
2TB
512GB, 1TB, 2TB
Internal Bus
Bandwidth per
Director
PCIe Gen3 x 80
lanes
(1 lane=1GB/s)
PCIe Gen3 x 80
lanes
(1 lane=1GB/s)
PCIe Gen3 x 80
lanes
(1 lane=1GB/s)

• SW designed to support 32 Logical Ports (Ports 0-31)
– Ports 0,1,2,3 in Slot 1 and 20,21,22,23 in Slot 7 are reserved
• Logical Ports numbered left to right bottom to top
– Across the 8 Slots that contain FE or BE I/O Modules
IO MODULES AND PORT NUMBERING
Slot
0
Slot
1
Slot
2
Slot
3
Slot
4
Slot
5
Slot
6
Slot
7
Slot
8
Slot
9
Slot
10
28
29
30
31
24
25
26
27
12
13
14
15
Vaultto
Flash
16
17
18
19
4
5
6
7
8
9
10
11
0
1
Managem
ent
A
Vaultto
Flash
Director
1 or A
Director
2 or B28
29
30
31
24
25
26
27
12
13
14
15
Vaultto
Flash
16
17
18
19
4
5
6
7
8
9
10
11
0
1
Managem
ent
B
Vaultto
Flash
Back-EndVault Flash Fabric
Management
Module
Vault FlashUniversal/FE Universal/FE
Vaultto
Flash
Vaultto
Flash
Vaultto
Flash
Vaultto
Flash

VAULT TO FLASH
Before• Reduced Environmental Cost
• Smaller Li-Ion SPS in every cabinet
only powers the directors
• Can have more drives per cabinet
• More usable capacity per raw TB
• Reduced space and weight
• Fewer cabs per TBu
• Improved performance and
Usability
• Larger memory space that can be
saved within 5 minutes
• Simplified configuration
After
Vault to
Flash

FRONT END IO MODULES
IO Module Interface Ports Speed
Host/
SRDF
8Gbps FC FC, SRDF 4 2/4/8 Gb/s
16Gbps FC FC* 4 4/8/16 Gb/s
SRDF
1Gb Ethernet
Copper/Optical
SRDF 2/2
2 x 10/100/1000 Mb/s (Cu)
and 2 x 1Gb/s (Opt)
10Gb Ethernet
Optical
SRDF
2 10 Gb/s
eNAS
10Gb Ethernet
Copper or Optical
eNAS - NFS/CIFS 2
1/10 Gb/s (Cu)
10Gb/s only (Opt)
1Gb Ethernet
Copper
eNAS - NFS/CIFS
4 1 Gb/s
Backup to Tape eNAS - NDMP 4 2/4/8 Gb/s FC
*16Gb FC for SRDF is a future

OTHER IO MODULES
IO Module Purpose
Flash Flash for Vault & metadata (175, 350, 700GB)
SIB Fabric connection
SAS Back End SAS IO Module

Lower Power & Cooling
45%
UP
TO
Improved
TB/ft.
4XUP
TO
75%
UP
TO
Reduced Footprint
Improved IOPS/tile
9XUP
TO
DUAL DENSITY, DRIVES DOWN TCO

EXPLOITING THE DENSE PACKAGING
• 10% Cap. 800GB EFD (3+1)
• 90% Cap. 1.2TB (6+2)
• ~50k IOPs/engine target workloads
– 85/15 or higher skew
– ~250 IOPs/TB or lower density
– ~2ms average response time
100K
• ~200 TB usable/engine (dual engine cabinet)
• ~$3/GB usable street price
– Including FAST Advanced Suite & 3 year SW maintenance
• Standard Rack size and 3rd party Racking
VMAX Sizer Output

• Drives are all SAS or NL-SAS
– No more FC or SATA drives
• All SAS drives are 2.5” media
– Also available in 3.5” carriers
– No more high power (but cheap) native 3.5” drives
• All NL-SAS (7k RPM) drives are 3.5” media
– Not available in 2.5” form factor
• Minimum of 4 drives per DAE (no more vault…)
• 1 spare for every 50 drives / engine required
DISK DRIVES

• Drives are attached to both directors
in a given engine
• All drives are SAS connected
• SAS drives are active/active
– Avoids failovers during IO Module
maintenance tasks
DRIVES AT GA
Type Capacities 2.5” 3.5”
SSD 200, 400, 800GB Y Y
15K RPM 300GB Y Y
10K RPM 300, 600, 1200GB Y Y
7K RPM 2, 4TB NO Y
For more info on maximum
drive limits per system, see
https://www.emc.com/auth/r
coll/salesguide/vmax3-hwsw-
ordering-guide.docx

PS (x2)
Fan(X3)
LCC(x2)
HDD(X60)
Up to
60TB
per
Rack U
DAE60 – 3.5” DRIVE CARRIERS

Fan(x10)
HDD(x120)
PS (x4)
LCC(x2)
Up to
200K IOPS*
per
Rack U
DAE120 – 2.5” DRIVE CARRIERS
* With Flash Drives

• All 240V Power
• All below power options supported for 100K, 200K
and 400K
• 3-phase delta – primary use case
– 2 drops/bay
• 3-phase wye – Country specific use cases
– 2 drops/bay
• Single-phase – Optional
– Up to 6 drops/bay
SYSTEM POWER OPTIONS

F
gfgsflg
gjridenvfn
kdkfm,fkdfk
qmd nvfnjvfjkl
dhwuidgbfthyelso
lfkfidemmfkfklfifmfm
dfjfnfndjdnjeismdw9459
0348945hjgfnfks8djd7a
jgjguiruirnfnfuie8jfkif
klfkfjkfirjwemdfdkidei
kfjgfheuiendfjdjshyeyu
vnfjdejcj j
nvjdjkjk
fmvfn
k
0
0010
110010110
001001001010101
1101011000110101110
10101010010000011101010
110011010100010001000111010
10101010011001010111111100000
101010010101001010010101010010
0010011001110011010010101010
10100101010101001000011101
101010101010101010101
1100110011001101
010100010110
1000100101
01010
1
VMAX3 D@RE ENABLED
• Hardware installed on all
systems
• Encrypts all user data on the
array - at the drive level
• All VMAX data services
supported
• External encryption key
managers supported
• Advanced Encryption Standard
(AES) - 256 encryption
• Will be submitted for FIPS-
140-2 validation
IN PLAN
2015

Dynamic Cores
OPTIMIZE PERFORMANCE AT SCALE
BALANCE RESOURCES
NO LONGER SILOED
RESPOND TO CHANGING
WORKLOADS
NO ADDITIONAL HARDWARE
PERFORMANCE
ISOLATION
PREDICTABLE SERVICE
DELIVERY

• Dual Redundant Infiniband
Switches
– High performance and
simplified cabling
– Carries control, metadata,
user data
– Speed – 56Gbps/port
– Matrices
• 18-port in 400K
• 12-port in 100K, 200K
– End point
• SIB IO Module
• One on each director
• Dual Redundant Ethernet
Switches
– Carries environmental
information
– Speed – 1Gbps/port
– Connects to every director
at its Management Module
– Always present, even in
single engine system
SYSTEM INTERNAL MATRIX CONNECTIONS

VIRTUAL MATRIX: VMAX 400K
Front of Rack
Rear of Rack
 Connects to every director
 18-ports - 1 port/director for
internal fabric
 Redundant, hot pluggable power
supplies
 Fan FRU
 Supported for dual 5-minute
vault periods by SPS

VIRTUAL MATRIX: VMAX 100K, 200K
 Connects to every director
 12-ports - 1 port/director for
internal fabric
 2, non FRU-able, power supplies
 Supported for dual 5-minute
vaults by SPS
 Same chip as 18-port Matrix
 Same management FW

Front-end (FA) ports
VMAX3
Default (balanced) setting
HYPERMAX
OS
Back-end (DA) ports
Front-end Core Pool
Back-end Core Pool
0 1
0 1 N-1 N
NN-1
MULTI-CORE TECHNOLOGY
Back-end (DA) ports
VMAX
0 1 N-1 N
0 1 N-1 N
• Single, dedicated core for each dual port
All cores
can be
applied
N-1
100%
load on
this port

Back-end (DA) ports
0 1 N-1 N
0 1 N-1 N
Back-end Core Pool
Front-end Core Pool
HYPERMAX
OS
MULTI-CORE TECHNOLOGY
Host Load setting Disk Load setting
Back-end (DA) ports
Front-end Core Pool
Back-end Core Pool
0 1 N-1 N
0 1 N-1 N
HYPERMAX
OS

• CPU cores are now always balanced
– Flash drives deliver 5+k IOPs each, CPU cores can fully
drive them
– No host ports that cannot be used effectively
– Do still spread host ports across directors
• Upgrade of DAEs added 1 at a time (Available 2015)
– 6GB/sec. bandwidth per director to each (direct connected)
DAE
– All CPU cores connect to all DAEs
– No need to balance drives within engines
• Pairs of Flash RAID groups per engine recommended
BALANCE HAS CHANGED

LOCAL RAID
• Improves BE performance
– Faster rebuild stays local on one
director (no matrix IO)
• No impact to availability
– Supports dual-initiator failover/failback
model as in VMAX
– Redundant access from each director to
each DAE
– Active on one director at a time (HA
failover)
– Striped across DAE power zones (4 per
DAE)
• Ease of: Adding new RAID groups with
new engines
– SRP spans all the engines and FAST will
rebalance capacity and load as engines
are added
All RAID group members now sit behind one engine
 All VMAX 1/2/400K use standard
rack sizes and supports 3rd party
Racks (there is a faster install
with an EMC Rack)
 Most flexible floor planning
capability in the industry
 Eliminates cross-bay cabling for
direct/daisy chain DAE cabling.
 Dispersion at engine/bay level
 New systems can be any
combination of contiguous or
dispersed bays
Core Local Raid benefits Additional local RAID benefits

(Single Engine System Bay)
VMAX 400K CONFIGURATIONS
System
Bay 1
System
Bay 2
System
Bay 4
System
Bay 3
System
Bay 8
System
Bay 6
System
Bay 7
System
Bay 5
Engin
e 1
Engin
e 2
Engin
e 3
Engin
e 4
Engin
e 5
Engin
e 6
Engin
e 7
Engin
e 8
Note: 100K scales to two System Bays, 200K scales to four System Bays, 400K scales to eight System Bays

(Dual Engine System Bay)
VMAX 400K CONFIGURATIONS
System
Bay 2
System
Bay 1
System
Bay 3
System
Bay 4

SINGLE ENGINE PER RACK
VMAX3 DISPERSED ARRAY
Engine 1
Engine
2
Engine
3
Engine
4
Engine
5
Engine
6
Engine
7
Engine
8
25 Meters from
System Bay 1

DUAL ENGINE PER RACK
VMAX3 DISPERSED ARRAY
25 Meters from
System Bay 1

HYPERMAX – GAME CHANGING OS
BUILT ON ENGINUITY
• 25+ YEAR HISTORY
• >1.5 BILLION RUN HOURS
NEW LEVELS OF TCO FOR VMAX
• SIMPLER MANAGEMENT AND HIGHER DENSITY
• INFRASTRUCTURE CONSOLIDATION
ENGINUITY + NEW CAPABILITIES

Rapid & Efficient Deployment of Data Services
EMBEDDED HYPERVISOR
Q4
Embedded Data Services*
• File
Q3
Embedded Application
Foundation
• Initial Support for Tools
– Symmetrix Traceability Tool
– Solutions Enabler
• Pre-installed To Improve Time
To Value
• Leverages VMAX HA
• Two VMs for High Availability
Careful of
futures
* Others to follow

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint

SIMPLE,
AUTOMATED,
WORKLOAD-
AWARE, SERVICE
LEVEL BASED
SIZING
PROACTIVE
PERFORMANCE
HEADROOM (PER
SERVICE LEVEL)
UPFRONT
ADMISSIBILITY
CHECK
VMAX Value – Simplify and ensure
consistent application responsiveness
• Performance specific service levels (e.g. ms.
response)
• Applies consistently across all iterations of
VMAX from now to eternity
• Aligns with enterprise customer service
architectures (e.g. ITIL)
• Hybrid Array message, always optimizing for
IOPs and capacity
– Ensures the best possible application of system
resources as workloads change. Simple
expansion. Continued leverage of all technologies
as tech vector changes and we shift to all Flash
CORE TECH VALUE PROPS- SLOS

Each Workload has a Unique Footprint and Cost
DIFFERENT WORKLOAD REQUIREMENTS
CAPACITY REQUIREMENT
TRANSACTIONREQUIREMNMET
DATABASE
FILE
VM’S
10 IOPS/GB
1 IOPS/GB
0.1 IOPS/GB

GATHER
CUSTOMER
PERFORMANCE
DATA
SLO Acquisition Model
CONFIGURE
RIGHT SIZED
SOLUTION
WITH SIZER
TOOL
100K
DIAMOND
GOLD
SILVER
BALANCED
SYSTEM
ORDER FULLY
CONFIGURED
IN FACTORY
SIMPLE
PROVISIONING
BY SERVICE
LEVEL

PROVISIONING FOR SERVICE LEVEL OBJECTIVES
What Was Complex
Create thin pools for FAST
Create TDAT devices
Set Pool Reserve Capacity
Create VP Tier using Pools
Create FAST Policy (% for each Tier)
Set FAST Relocation Rate
Select protection (RAID)
Select underlying disk technology
Set FAST Time Windows
Is Now Simple
Create storage group (name)
Select cascaded parent / child / neither
Choose/create a host
Associate child SGs to parent
Set device capacity & count
Associate SG to FAST Policy
Build meta configurations as needed
Add volumes to storage group
Choose port group
Bind devices to thin pool
Choose SLO

Virtual
Pools
Storage
Resource
Pools
Storage
Groups
Physical
Disk
Groups
FAST ELEMENTS
VP_ProdApp3VP_ProdApp1
Pool 0
RAID 5
(7+1)
Pool 1
RAID 1
Pool 2
RAID 5
(3+1)
Pool 3
RAID 6
(6+2)
DG 0
eMLC
200GB
DG 2
10K
600GB
DG 3
7.2K
2TB
DG 1
15K
300GB
VP_ProdApp2
Storage Resource Pool 0 (Default)

Turning Complexity into Simplicity
STORAGE RESOURCE POOL
203.1
203.2
203.3
203.4
203.5
203.6
202.1
202.2
202.3
202.4
202.5
202.6
201.1
201.2
201.3
201.4
201.5
201.6
200.1
200.2
200.3
200.4
200.5
200.6
TDAT VIRTUAL THIN POOL 0 – EFD RAID5 (3+1) TDAT VIRTUAL THIN POOL 1 – 7.2K RAID6 (6+2)
ENGINE 1 ENGINE 2
100.24
100.3
100.2
100.1
TDAT 100
101.24
101.3
101.2
101.1
102.24
102.3
102.2
102.1
103.24
103.3
103.2
103.1
TDAT 200 TDAT 201 TDAT 202 TDAT 203TDAT 101 TDAT 102 TDAT 103
RAID5 (3+1) RAID6 (6+2)
103.P 103.1 103.2 103.3
102.1 102.P 102.2 102.3
101.1 101.2 101.P 101.3
100.1 100.2 100.3 100.P
203.1 203.2 203.3 203.P
202.1 202.2 202.3 202.4
201.1 201.2 201.3 201.4
200.1 200.2 200.3 200.4
`
203.Q 203.4 203.5 203.6
202.P 202.Q 202.5 202.6
201.5 201.P 201.Q 201.6
200.5 200.6 200.P 200.Q
PHYSICAL DISK GROUP 0 – EFD 200GB
TDAT100 LOGICAL SPLITS ARE 100.1 THRU 100.P
PHYSICAL DISK GROUP 1 - 7.2K 4TB
TDAT200 LOGICAL SPLITS ARE 200.1 THRU 200.Q (UP TO 24 SPLITS ALLOWED PER PHYSICAL DISK)
D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12
203.24202.24201.24200.24
FAST PROMOTES &
DEMOTES DATA IN 42
TRACK SUPER-SLICES
STORAGE RESOURCE POOL 0 (DEFAULT)
Gold Storage Group
Bronze Storage Group Diamond Storage Group

• The Storage Resource Pool (SRP) is a single container that contains
hybrid storage resources
– Ensures ease of management
– Abstracts all the configuration complexity of typical arrays
• The SRP is a collection of data pools that make up a FAST Domain
– A data pool can only be included in one SRP
• SRPs are defined within the bin file and not configurable by the
customer
• The factory preconfigured array includes one SRP that contains all
data pools in the array
• Multiple SRPs may be configured on site (we do not recommend)
– One SRP must be marked as the default
– For stringent (e.g. legal) physical data separation use cases
STORAGE RESOURCE POOL

FAST: THE FOUNDATION FOR SLO!
• Capacity based
• Manual policy settings require
monitoring
• No specific SLO (ms,. IOPs)
• No guarantees
• No system level impact
performance awareness
• Slice prioritization based on
heat (# of IOPs)
FAST – VMAX2
Tier 3
1 TB
SATA
RAID 6
(14+2)
Storage Tier
Tier 1
200 GB
Flash
RAID 5
(3+1)
Tier 2
146 GB
15K FC
RAID 1
Storage Groups
Tier 4
FTS Per
vendor
FAST Policies
5%
25%
100%
100%
100%
Platinum
Silver
20%
Bronze
100%
VMAX

• Next Gen FAST
– The engine that
powers SLO
– Advanced metering
• Service Level
Objective
– Defines a specific
performance range
– Applicable to
Storage Groups
FAST: THE FOUNDATION FOR SLO!
FAST VMAX3: Proven VMAX technology, now to the power of 3
Disk
Group 3
SAS 1.2
10K
R5 3+1
Storage Resource
Pool
Disk
Group 1
EFD 400
R5 3+1
Disk
Group 2
SAS 300
15K
R1
Storage Groups
Disk
Group 3
NL-SAS 4T
7K
R6 6+2
ServiceLevelObjectives
Diamond
SLO
Platinum
SLO
Bronze
SLO
Gold
SLO
Silver
SLO
Optimized
BASIC only option for FAST Base
VMAX3
NEWInternalFASTProcess

• Greater effective use of high cost resources
– Smallest sublun granularity in the Industry (128KB)
• More Responsive to workload changes
– Collection and analysis time interval of 5 mins
• Enables Intuitive, Achievable Service Levels:
– FAST intimately understands the workload being placed on the array
• Metering for: Reads, Writes, Sequential Reads, and IO Sizes
– FAST knows if a workload is cache un-friendly
• Based on Cache hits and if needs be moves on to faster back end storage
– FAST understands high write workloads and aligns with best raid type
• IO clustering, Based on r/w ratio within the SRP
– FAST understands the skew and locality characteristics of workloads
• So it can be efficient in terms of assigning the most optimal tiers of storage to meet
the performance need
– FAST intimately understands the load on the whole system
• Tracks at Disk Group and Storage Group level as well as sub LUN
NEW FAST ENGINE
WHAT IT MEANS

• FAST collects performance metrics continuously at three levels
– Disk group/virtual pool, storage group, sublun
• The sublun tracking granularity at three levels, smaller than VMAX
– Extent: single track (128 KB) vs. 12 tracks (768 KB)
– Extent Group: 42 extents (5.25 MB) vs. 10 extents (7.5 MB)
– Extent Group Set: 42 extent groups (220.5 MB) vs. 48 extent groups (360 MB)
• Performance metrics collected include
– Reads, writes, sequential reads, cache hits
– IO size
• Tracked separately for read and write
– IO clustering
• Based on r/w ratio within the SRP
• Influences where to move workload types (write alignment)
– Time since last accessed
NEW FAST ENGINE
PERFORMANCE METRICS

• The Service Level Objective (SLO) defines the ideal
performance operating range of an application
• Each SLO contains an expected average response
time value
– Actual response time experienced by SGs will depend on
• IO size (small vs. large)
• IO type and mix (read vs. write; hit vs. miss)
• Replication (local vs. remote; sync vs. async; replication distance)
– The SLO can be combined with a workload type to further refine the objective
– SLOs can be changed on the fly at any time as service needs change
• SLOs are predefined and come prepackaged with the array
– Not customizable by the customer
• The default SLO is Optimized
– Protects disks from being overloaded
– Places the most active data on the highest performing storage and the least
active on the most cost-effective storage
SERVICE LEVEL OBJECTIVE
Define specific performance ranges for your workloads
SLO is not a
guarantee !!

WORKLOAD TYPE
PRE-DEFINED SERVICE LEVELS

WORKLOAD TYPES
Workload Description
OLTP Small block IO workload
OLTP_Rep
Small block IO workload
with local and/or remote replication
DSS Large block IO workload
DSS_Rep
Large block IO workload
with local and/or remote replication
None
No pre-definition of workload type – The system will
do the best it can*
* If you are not sure of the workload, specify None. The system has knowledge of IO type.
This is better than guessing and being wrong

SERVICE LEVELS
PRE-DEFINED SERVICE LEVELS

SLO DEFINITIONS
SLO Name Description
Diamond Emulates EFD performance
Platinum Emulates performance between EFD and 15K RPM drive
Gold Emulates 15K RPM performance
Silver Emulates 10K RPM performance
Bronze Emulates 7.2K RPM performance
Optimized
(Default)
Does the best it can after it has satisfied the above SLO’s
(Achieves optimal performance by placing most active data on
higher performing storage and least active data on most cost-
effective storage)

APPROXIMATE AVERAGE RESPONSE TIME
SLO Name OLTP OLTP_Rep DSS DSS_Rep None
Diamond 0.8 2.3 2.3 3.7 0.8
Platinum 3.0 4.4 4.4 5.9 3.0
Gold 5.0 6.5 6.5 7.9 5.0
Silver 8.0 9.5 9.5 10.9 8.0
Bronze 14.0 15.5 15.5 16.9 14.0
Optimized
(Default)
N/A N/A N/A N/A N/A
*all response times shown in milliseconds (ms)
* Actual response time experienced by SGs will depend on IO size, IO type and mix, and replication options

CAPACITY COMPLIANCE
SLO Name EFD 15K 10K 7.2K
Diamond Y N N N
Platinum Y Y Y N
Gold Y Y Y Y
Silver Y Y Y Y
Bronze N Y Y Y (Required)
Optimized
(Default)
Y Y Y Y
*When pools are full, to avoid failed allocations, the data may be placed in pools marked “N”

Architectural Components that Deliver SLO’s
FAST IMPLEMENTATION
Internal FAST Unisphere for VMAX
HYPERMAX OS Management
Solutions Enabler SLO
External FAST
Pattern Recognition
TOOLS GuestOS
External Hint
TranslationWorkload Forecasting
Data Placement
Performance Metrics
Workload Planning
Workload Classifying
Array Advising
Base FAST FoundationAdvanced FAST
2015

Simple, Automated, Workload-Aware, Service Level Based Sizing
PLAN FOR SERVICE LEVEL- UNISPHERE
Workload Planner
Features Supported by WLP
• VMAX Sizer - Capacity planning
• Perf. Headroom indicator -
Remaining capacity
• Admissibility Check –
Provisioning

• Runs within the Tools GuestOS
• Two GuestOS instances provide high
availability
– Data is shared between both instances
• Performs analysis to enable long-term
workload regularity/pattern detection
– Assists in improving FAST workload
forecasts to achieve SLO compliance
• Provides translation services for
application hinting
– Will support out-of-band hints from
external hint providers
EXTERNAL FAST
External FAST
Pattern Recognition
TOOLS GuestOS
External Hint
Translation
2015

DATA PLACEMENT INTELLIGENCE ALGORITHM
FAST data movement generated by algorithm
• For each virtual pool and each FAST managed SG
• Takes into account observed workload
Determines workload forecasts
• To adjust the load distribution on each pool
• To achieve SG SLO compliance
Generates a list of best extent groups to move

• FAST data movements are generated by the
data placement intelligence algorithms
• The algorithms focus on
– Capacity compliance
– Disk resource protection
– SLO response time compliance
• Generate a list of the best extent groups
(5.25MB) to move
– To adjust the load distribution on each pool and
– To achieve SG SLO compliance
– Once we have met the SLO we stop moving for
that SG
DATA MOVEMENT

ALLOCATION AND REBALANCING
Allocation Management
• Allocations can be made in
any virtual pool within the
SRP
• Pools are ranked according to
their ability to handle writes
• Allocation is performed
according to the pool rank in
the SRP and the device SLO
Disk Group
Ranking for
Incoming Writes
(Diamond, Platinum, Gold,
Bronze, Optimized)
Raid-1, 15K
P, G, B, O
Raid-5, 10K
P, G, B, O
Raid-5 EFD
D, P, G, O
Raid-6, 7.2K
P, G, B, O

• Aims to protect drives and data pools from
being overloaded
– Especially avoid overloading 7.2K
• Places data on the most appropriate media
(drive technology and RAID protection)
– Heavy read workload moved to EFD
– Heavy write workload placed on RAID 1 (15K
or 10K)
– No activity or very low activity to 7.2K
• Provides basis for the default Optimized SLO
DISK RESOURCE PROTECTION

Constantly working to balance load and ensure optimal operations
DISK RESOURCE PROTECTION ALGORITHM
+IO
- IO
+ GB- GB
Accept capacity
Accept IOPs
Accept capacity
Remove IOPs
Remove capacity
Accept IOPs
Remove capacity
Remove IOPs

EASILY MONITOR SERVICE LEVELS
STORAGE DASHBOARD

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint
IN PLAN Q4
Local Replication

VMAX Value – massively reduced cost
• Data copies are endemic
– Most data duplication comes from wholesale copies of a consistent app environment
e.g. prod copies for test, dev., QA etc
• Full clones are impractical (Often 10x capacity)
• Advanced local replication technologies will differentiate our offering
– Low cost (pointer based, no copy resources required)
– Scaling – we can make 1000’s of copies
– Simple to take and restore
– Crash Consistent or application consistent with AppSync
– High Speed
– Don’t dedupe, predupe
• Many dedupe technologies don’t deal well with offset based dedupe
CORE TECH VALUE PROPS- LOCAL REPLICATION
Pre-Dupe rather than dedupe

Production
Volume
Linked
Target
w Minimal Cache
w Pointers not Devices
w Best Performance
Snaps as Versions
User-assigned Names
Storage Group Operation
256 Snaps
1024 Linked Targets
Automatic Expiration
Linked
Target
8 AM Snapshot
9 AM Snapshot
10 AM Snapshot
11 AM Snapshot
12 PM Snapshot
1 PM Snapshot
2 PM Snapshot
3 PM Snapshot
4 PM Snapshot
5 PM Snapshot
Start of Day
Noon
End of Day
w Copy and Restore
EASILY MANAGE MANY COPIES WITH SnapVX

Common launch page & simple SnapVX setup
APPLICATION CENTRIC PROTECTION WIZARD
Select Snap Option
Add Snap
Name Option to
enter a new
Storage
Group that
Snap will be
associated
with

• Scale - Support Next Generation Needs
– Supports up to 256 Snaps/source (Architected for 1K
Snaps/Source)
– 1024 linked targets/source
– Unlimited cascaded hops from every source
– Delete out of order
• Target less Snapshots - Efficientusage of system resources
– Extreme Cache Efficiency
– N-1 times cache savings on N Snaps
– Unlinked Snaps do not increase device count
– Fast Create and Terminate Times
• Performance Enhancements – Zero Impact Snaps
– No Impact to production device when taking snaps
– Redirect on Write Technology
TIMEFINDER(TF) SnapVX BENEFITS

• Predupe – Better than Dedupe
– Zero impact Snaps facilitates moving from
multiple copies of Clones to Snaps
– Creates significant space savings
• Management – Ease of Use
– Named Snapshots
– Expiration Dates
– Storage Group Level Snap shot
TF SnapVX BENEFITS CONT.

• TF SnapVX 256 Snapshots
– Max 1024 linked to a single source in a chain
– TF Snap VX should be used instead of emulated TF due
to increased scale, performance and efficiency
• Legacy TimeFinder Options: VP Snap, Clone and
Mirror
– Primarily to support existing scripts
– Legacy TimeFinder options are emulated by TF SnapVX
– Old TimeFinder Limits apply
• Limited to 32 TF VP Snaps
• Limited to 16 TF Clones
• Requires that all emulations consume cache
• SRDF interoperability: with SnapVX, Clones and VP Snap
TF RELEASE FEATURES

TDAT TDAT TDATTDAT
and Snapshot Delta
REDIRECT-ON-WRITE
Source Snapshot
Snapshot Delta
(Point-in-Time)
New Write 1.0 Original Track1.1

DEFINING MECHANISM
SNAP 1 SNAP 2
TDAT
Devices in
Storage
Resource
Pool
Source Volume
SNAPSHOTS
SNAP x
... Linked
Target
Tracks on Linked Target Being Defined

Target 1
CASCADING SNAPSHOT FROM A LINKED TARGET
Linked
Target
Source
Target 2
……
Target 3
Target X
Defined
(copy or nocopy)
Snapshot
Snapshot

STORAGE GROUP SNAPSHOTS
SRC 1
SRC 2
SRC 3
Source Storage Group
TGT1
TGT2
TGT3
Target Storage Group
Snapshot
Snapshot
Snapshot

• TF Clone/TF Mirror/TF Snap VP (Emulation Mode
Support)
– Existing scripts will work seamlessly
– These will be supported in emulation mode and behave
similar to current generation VMAX
– Be aware that you gain scale and flexibility by using SnapVX
• TF Snap (Deprecated)
– Existing scripts will need to be modified
– EMC provides TF scripting services to help with transition
and facilitate reaping the phenomenal benefits of SnapVX.
TRANSITION TO TF SNAP VX

VMAX3 Architecture
Advanced Availability-SRDF
Local Replication
Embedded File
Management
ProtectPoint
IN PLAN Q4
Remote Replication

• VMAX Value – Best DR solution in the industry
– Scale you DR as your entire IT infrastructure scales
• Existing customers running 100’000’s of LUNs being replicated
– SRDF/A Multi-Session Consistency replicates many
arrays with full consistency
• Applies to cloud scale deployments
• Applies to modern 2nd platform federated data solutions (e.g.
mainframe and open systems federated systems.
– VPLEX delivers on continuous ops even in the face of a
disaster
• 7x9’s of availability
• Active/Active metro area solutions truly ride through whole site
losses
• Also enables transparent data migrations , so no more planned or
unplanned down time
CORE TECH VALUE PROPS- NON-STOP OPS
INDUSTRIES MOST ROBUST ADVANCED AVAILABILITY OPTIONS

SRDF ARCHITECTURE
PROVEN IN THE LARGEST ENTERPRISES ON THE PLANET
• SRDF continues to be
the gold standard for
remote replication
– Offering Zero RPO for
Syncronous
environments
– And Near Zero RPO for
Async environments
• At Q4 GA Enhanced
SRDF will support
– 2 site replication
between VMAX3 Arrays
– 2 Site Replication
between VMAX and
VMAX3 arrays
– Concurrent replication
between VMAX3 arrays or
a mix of V2 and V3
Concurrent2 Site Configurations Supporting
VMAX arrays running 5876
VMAX

SRDF KEY RELEASE FEATURES
Supported in Q4
• SRDF Sync/Async/Adaptive copy (disk mode) 2 site:
– VMAX2-VMAX3 (N-X support)
– VMAX3-VMAX3
• Concurrent SRDF (3 site)
• SRDF HW Compression GIGE/FC
• SRDF SW Compression GIGE/FC

SRDF INTEROPERABILITY
SRDF/FAST Behavior
• SRDF devices can be associated to Storage group with any
of the pre-defined SLOs
• Customers can chose to have Remote SRDF devices on a
different SLO compared to Source side
• If no SLO is specified on Remote side, it will default to
“Optimized” SLO
• FAST SRDF Co-ordination between VMAX3 and VMAX2 will
not be supported in Q4

SRDF REMOTE SELECTION
APPLICATION CENTRIC PROTECTION WIZARD
Select remote Symmetrix
Remote 5876 asks for Thin
Pool and FAST Policy
Remote VMAX3 asks for SL
Options are TP & FP or SL,
never both

ENHANCED SRDF - WHAT’SNEW
 SRDF/A Multi-Cycle Mode (MCM)
 DSE to Storage Resource Pool (SRP)
 SRDF/A Write Pacing to DSE Spill Rate
 SRDF Groups per Director increased 64->250
 SRDF Dynamic Port Emulation
 SRDF N-X Backward Compatibility Support
 SRDF HW Compression Support
 Remote SID Availability w/Partitioned Pair State
 SRDF/S Performance Improvements
 128K Track Size for SRDF paired devices
Bolded items are
new features of SRDF

NEWSRDF/A MULTI-CYCLEMODE(MCM)
 Allows more than two Delta Sets on Primary Site
 Cycle are More Granular with Incremental Updates
 Cycle switch occurs at Min Cycle Time, regardless
 RPO at target is now more Granular & Predictable
Secondary SymmetrixPrimary Symmetrix
WAN N-M
R2
N-M-1N
. . .
N-1 N-M

NEW: SRDF N-X BACKWARDCOMPATIBILITY
 SRDF will be allowed from existing VMAX (V1/V2) to VMAX3
systems running Enginuity level 5977
– Requires SE 8.0, previous versions of SE will not be able to
discover a VMAX3
– DMX Support post-GA (Enginuity 5773)
 Many SRDF controls will not be allowed on SRDF N-X pairs
with pre SE 8.0 because the SRDF pair state will be Partitioned

NEW: SRDF/S PERF. IMPROVEMENTS
 Pipelining increases performance on large writes by removing current
requirement to wait every 8 tracks
 Reduces Global Memory (GM) access by including additional
information in the HA -> RA messaging previously read from GM
 HA now sends an RA pre-request when it starts a J0 write. RA then
creates job and prepares to start the IO as soon as the request
arrives

NEW128 KB TRACK SIZE SUPPORT
 VMAX3 will support multiple SRDF protocols depending on remote
track size
– Approach is similar to how we addressed previous VMAX->DMX
(64 KB->32 KB) support
 2 tracks on Legacy map to 1 track on VMAX3
 Doubling sectors, retaining sector size
 Outgoing 128 KB IO split into two 64KB IOs
 5876/5773 initiators – all IOs are 64 KB only
 No support for 32 KB (5671) legacy systems

OPEN REPLICATOR - WHAT’S NEW
 Open Replicator for Symmetrix (ORS) has been re-architected on the
VMAX3 platform to improve HA performance and relax the front end
pathing requirements
 Accomplished by separating the target IO path from the initiator IO
path via support for new Enginuity Data Service (EDS) emulation
 EDS emulation support has resulted in changes to ORS functionality
in several key areas

• Supported in Q4
– Hot Pull with Donor Update
– Cold Pull
– Front End Zero Detection
– Ceiling
– ORS with PPME
• Q4 Qualification Limitations
– Max number of ORS sessions per VMAX3 = 512
ORS- ANOTHER REPLICATION OPTION

VMAX3 MIGRATION OPTIONS
VMAX3
Migration Options
Technology
At First Availability (Q3) Q4 release
Source Array Source Array
VMAX2 DMX Competitor VMAX2 DMX Competitor
Open Replicator Push
(from VMAX2/DMX) Yes Yes X Yes Yes X
Open Replicator Pull
(from VMAX3) X X X Yes Yes Yes
VPLEX Yes Yes Yes Yes Yes Yes
SRDF X X X Yes X X
PPME Yes Yes Yes Yes Yes Yes
LVM tools (Host Based) Yes Yes Yes Yes Yes Yes
Open Migrator* Yes Yes Yes Yes Yes Yes
* NOTE: Open Migrator is only available for Windows OS. (Unix/Linux is no longer available or supported)

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint
IN PLAN Q4
Embedded File

• EMC VMAX eNAS – First Fully integrated High end NAS
solutions in the industry
– VMAX3 fully supports running mission critical VMware and Oracle
workloads on eNAS
– Robust, proven NAS stack (software taken from VNX File OE)
– Extends the value of VMAX3 and its features to NAS
– Simple integrated management (VMAX Unisphere File Dashboard)
– Market leading functionality
– Scale out technology
– More cost effective solution
• Less hardware, smaller footprint, faster “time to first data”.
CORE TECH VALUE PROPS- EMBEDDED FILE
THE ONLY “DESIGNED FOR MISSION CRITICAL WORKLOADS” NAS SOLUTIONS

• The Only NAS for VMAX3
• Virtual instances
– Of VNX2 NAS software
– No physical gateways
– Software data services
• Support for VMAX3 data services
as they are offered:
– SLO Provisioning
– FAST VP
– Host IO Limits
DATA SERVICES
HYPERMAX OS AND EMBEDDED NAS
(eNAS)
IN PLAN
Q4

eNAS OVERVIEW
• Optional selection when configuring a VMAX3
• 2 Control Stations and 2 Data Mover VMs are the minimum
configuration
• Data Mover attributes differ per VMAX3 platform
– # of Data Mover VMs, logical cores, memory, IO slots
• SW is pre-installed in Mfg  “network ready” upon reaching the
customer

CONFIGURATION HIGHLIGHTS
• VMAX 100K
– 2 Data Movers, 2 Control Stations, up to 2 IO Cards per Data
Mover instance, 6GB physical memory per DM
• VMAX 200K/400K
– 2 or 4 Data Movers, 2 Control Stations, up to 3 IO Cards per
Data Mover instance, 24GB physical memory per DM
• IO Cards consist of:
– 4 port 1GbE BaseT
– 2 port 10GbE optical
– 2 port 10GbE BaseT
– 4-port 8Gb FC (for backup to tape option only)

FILE
GATEWAYS
VMAX3 eNAS Overview
LESS EXPENSIVE TO
DEPLOY
UP TO 33%
HIGH END BLOCK &
FILE
CONSOLIDATION
REDUCED POWER
AND FOOTPRINT
UP TO 30%
EMBEDDED
FILE SERVICES
REDUCES TCO

VMAX HYPERVISOR CONCEPTS
• Provides a “virtual” machine host
platform DSH
VM– CPU Processing
▪ Symm/K thread per vCPU
– RAM
▪ From mirrored director memory
– Data storage for boot and
application data
▪ Cut-through device acts like an HBA
▪ Data storage is provided by LUNs
from the array
– Network Interface
▪ Enables client network access in and
out of the virtual machine
LUNs
CTD
vCPU RAM vNIC

FA Emulation
VMAX HYPERVISOR CONCEPTS
• The CTD has two components to enable access to
LUNs through an FA port
– CTD Server thread
• Runs on the FA emulation as a Symm/K thread
• Communicates with CTD client in the Data Mover VM
– CTD Client driver
• Embedded in VM host operating system
• Communicates with CTD server thread on FA emulation
• An operating system running in a VM must have
the CTD client driver installed to “see” storage
CUT-THROUGH DEVICE [CTD]
VM
port
VM
port
CTD
Server
CTD
Client

eNAS SYSTEM ARCHITECTURE
FA
NAT 1A
v
N
I
C
v
N
I
C
C
T
D
GOS BMC
DM3 Guest
(standby)
CS0 Guest
(primary)
GOS BMC
C
T
D
IP SLICs FC SLIC
File Protocol Clients,
CAVA, Quota Manager etc
Tape
Library
Director
Board 1
Control LUNs Data LUNsBoot LUNs/
Shared LUN
FA
NAT 2A
v
N
I
C
v
N
I
C
C
T
D
GOS BMC
DM2 Guest
(primary)
CS1 Guest
(secondary)
GOS BMC C
T
D
IP SLICs FC SLIC
Director
Board 2
MMCS
MMCS
File Protocol Clients,
CAVA, Quota Manager etc
Optional Tape
Library
Control
station
External
clients
Enginuity
Internal
Network
PCI Passthrough
PCI Passthrough

Enas PLATFORM CONFIG SPECS
• Data Movers are added in pairs
• The disk space required is the same for each Platform (~ 600GB per NAS System)
• eNAS always leaves at least two IO Module slots per engine for “block” IO for standard
VMAX operations
VMAX 100K VMAX 200K VMAX 400K
Data Mover VMs (2) Logical Cores 8 10 16
Memory (GB) 12 48 48
IO Modules ≤ 4 ≤ 6 ≤ 6
Control Station VMs (2) Logical Cores 2 2 2
Memory (GB) 8 8 8
SLICs None Req’d None Req’d None Req’d
Max DMs supported 2 4 4

VMAX eNAS VALUE STATEMENTS
• Predictable, scale-able
performance at load – easily serve
>6,000 active SMB connections
• Meta-data logging file system –
(mostly) rapid (few seconds) file
system checks on reboot, and no
file system fragmentation issues
• Small attack surface – not
vulnerable to viruses targeted at
general purpose OS's. No ability
to install 3rd party software. No
login.
 Fully integrated, cost effective
 Off-board, scale-out Anti-Virus
 Built-in Async File level remote
replication (File Replicator)
 Built-in NDMP
 Flexible and secure multi-
protocol file sharing
 Multiple file server identities
(CIFS and NFS servers) - file
server consolidation/multi-
tenancy
 No "Patch Tuesday"

eNAS BUILT-IN FILE FUNCTIONALITY
pNFS Support
Compression and
Deduplication
Up to
50%savings
File-level deduplication and compression
Compress
and dedupe
~ 500 GB
File data
Active Inactive
1 TB
File system Replication
Remote protection, distribution and DR
Production Local Site Remote Site
10-minute RPO 2-hour RPO
LAN WAN
Network
Snaps
FS/
LUN
Network
Snaps
FS/
LUN
Network
Snaps
FS/
LUN
File-Level Retention
Write Once-Read Many (WORM)
Common Event Enabler
Integration with third-party anti-
virus and quota management vendors
Anti-virus/QM
serverUser
VMAX
Improved performance and scalability
Application
servers
Fiber
Channel
or iSCSI
IP
NAS
request
SAN
Delivery
Cloud Tiering Appliance
Integration with archive engines
for tiering to the cloud
Public
Cloud

UNISPHERE FOR VMAX FOR FILE
PROVISION
FILE
SYSTEMS
DISPLAY
STORAGE
MAPPING
MONITOR
BLOCK & FILE
RESOURCE
CREATE STORAGE
POOLS

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint
Management

• Cost of storage operations significantly outweighs the cost
of storage purchase
• Intuitive, integrated, context specific management
through Unisphere and our integrations
– EMC number 1 for virtual environments 3 years running
• Maximum ease of use
– SLOs change the game
– 5 clicks to provision and protect
– Wizards
– Dashboards
– Integrated File management
CORE TECH VALUE PROPS- MANAGEMENT
ADVANCED FUNCTIONALITY, YET SIMPLE TO MANAGE AT SCALE

ADVANCED
DATA SERVICES
SIMPLE
DATA SERVICES
Transaction
Processing
Online
Archive
VM
Infrastructure
Decision
Support
Full Motion
Video
High
Performance
Computing
Web
Content
Content
Management
IT
Applications
Storage
As-A-Service
Home
Directories
Email
Application
Development
Tech
Apps
FLASH
15K RPM
FC/SAS
10K RPM
FC/SAS
7K RPM
SATA
WHAT IF YOU COULD MANAGE
BY SERVICE LEVEL
VIA POLICY BASED
APPLICATIONS
OBJECTIVES
AUTOMATION
WORKLOAD MANAGEMENT CHALLENGES
EMC makes no representation and undertakes no obligations with regard to product planning information, anticipated product characteristics, performance specifications, or anticipated release dates
(collectively, “Roadmap Information”). Roadmap Information is provided by EMC as an accommodation to the recipient solely for purposes of discussion and without intending to be bound thereby.

UNISPHERE MANAGEMENT FAMILY
Consistent and Efficient
VNX VMAX
RECOVER
POINT VPLEX
CONTEXTUAL
MANAGEMENT
EASY TO USE
TASK BAR
PROVISION
IN SECONDS

UX REDESIGN FOR SL-DRIVEN SIMPLICITY
KEY CHANGES TO UNISPHERE FOR VMAX 8.0
Benefits
• Higher operational efficiency day to day tasks
• Faster ramp up for new admin
Features
• Dashboards - Storage, performance,
protection…
• Wizards - Provisioning, protection…
• Cleaner “Look and feel”

New tabs
• Navigation can be done by
moving between the 3 tabs
– Monitor
– Analyze
– Charts
• Creating charts would be
done in a dedicated tab
UPA- USABILITY AND NAVIGATION

UPA - TCE enhancements
• Complete UI refresh
• A dedicated area for
creating charts
• Sending generated
reports via email
• Export charts in 1 click
• Timezone selection for
the scheduled reports
• Dynamic categorization
in the analyze section

DASHBOARDS – QUICK ACCESS TO INFORMATION

UPA: UTILIZATION DASHBOARD
New tabs
• Scalable
• Quickly identify which
component is overloaded
• The user can then expand
items of interest to see the
individual components
• Balance
• based on the difference
between busiest and least busy
component measured against
threshold
• Queue Depth utilization
• A new metric (similar as the
queue depth in Symmerge)

• Event correlation
• Correlate SG response time with high
utilized hardware component
• A dialog box showing a chart with SG
response time and %busy for the
hardware component will be open
when hovering the mouse on the
red/yellow icon
• Click on a given storage group
opens up the single SG
dashboard
UPA: STORAGE GROUP DASHBOARD
Highlight performance problems on the SG level
Potential root
causes
SGs not
meeting SLOs

UPA: HEATMAP
V2 V3

• Unisphere 8.0 introduces a
database awareness solution –
Database Storage Analyzer (DSA
– previously known as DBclassify)
• DSA is targeted for Oracle
databases running on VMAX
providing the following key
values:
• Reduce Performance
troubleshooting time
• Accelerating performance of
mission critical processes
UNISPHERE - DATABASE STORAGE ANALYZER

• Unified view across
DB & storage
• Reduce troubleshooting
time
• Identify DB bottlenecks
that are not related to
the storage
• Show the impact of
FAST VP on DB
performance
DATABASE PERFORMANCE DASHBOARD
TROUBLESHOOT DB-STORAGE PERFORMANCE ISSUES
Compare DB and Storage
response time

• Storage response time is good while in the DB it looks like an IO issue
• DB “PROD-1” suffers from high IO response during the day
• DSA shows that the storage response time is good (much lower than
the DB response time) but the OS CPU load is high
USE CASE 1

• Storage response time is bad as a result of load on the SATA drives
• DB “PROD-2” suffers from high IO response time on Tuesday at 08:30.
• Drill down to DSA performance tab shows that most of the DB IOs are
indeed coming from SATA drives.
USE CASE 2

CORRELATINGTHE DB IOTO SYMM DEVICES
Shows how much time users are waiting for
each Symmetrix device that is associated
with the database

Identify top objects in terms
of IO wait
IO Type: random Vs.
Sequential
ANALYTICS: DEEP DIVE IO ANALYSIS

APIS FOR AUTOMATION & SIMPLIFICATION
Features
• Provisioning (legacy & SLO based VMAX3)
• Replication
• Planning
• Performance and Health Monitoring
• APIs for creating Storage Groups using
predefined service levels
• APIs for Headroom indicator
Admissibility
Check
SL
Provisioning
CheckAdmissibility
Lookup (workload,…)
createStorageGroup(
GroupName, SLO, …)
Benefits
• Provide simple to use APIs that provide
higher level of abstraction for simpler
operational model
• Users or partners can create a custom
workflow using REST APIs to automate
SLO based storage provisioning.

ORACLE ENTERPRISE MANAGER 12C INTEGRATION
INCREASE DBA PRODUCTIVITY
 Set of tools for managing the
Oracle environment
 Manage/monitor the Oracle
environment including storage
 Automate tasks and assist on
database administration
 Monitor VMAX utilization,
configuration, and performance
 Set thresholds and notifications
for key VMAX health indicators

VMAX3 Architecture
Remote Replication
Local Replication
Embedded File
Management
ProtectPoint
IN PLAN Q4
ProtectPoint

DATA PROTECTION CHALLENGES
• More stringent SLAs with exponential data growth
• Traditional backup and/or snapshots not meeting requirements
Bridging the Gap in Data Protection Solutions Today
SnapshotReplicationAvailability Backup Archive
Availability Durability
NOTE: Most data is protected in multiple ways for multiple reasons

Secondary/DR SitePrimary Site
Primary Storage with
Local Snapshots
Protection Storage
with Local Backups
Primary Storage with
Replicated Snapshots
Protection Storage with
Replicated Backups
Protection
Availability
Protection of First and Last Resort
DEPLOYMENT OF THE CONTINUUM
Application
Owner
Backup
Admin
Storage
Admin
EMSD
SnapVX/SRDF
Use Case
DPAD
ProtectPoint
Use Case

• Application Impacted
– All data passes through
application server
• Missing SLAs (RPO/RTO)
– Too much data to move
– Too complex a data path
• Complex and Expensive
– Infrastructure required to
meet SLA continues to grow
TRADITIONAL BACKUP IS (STILL) BROKEN
Even though EMC DPAD has done a lot to improve things…
Production
Backup
Agent
Primary Storage Protection Storage
Backup
Server
Application
Server
Backup
Admin

• Snapshots have their place as does support for
snapshots within backup software
– Recovery of First Resort
– Short term retention of limited set of copies
– Creation of Test & Dev. copies of production apps
• Layering software on top of snapshot doesn’t
change the fundamentals of snapshot
– Reliability: corruption cascades through the set
– Scale: typically limited to 250 snapshots/copies
– Recovery: complex recovery processes
– Replication: introduces more snapshot challenges
– Storage efficiency: limited and to snapshot chain
– Performance: may impact production
– Application integration: usually limited
WHY IS SNAP & REPLICATE IS NOT ENOUGH ?
Even though it has value outside of backup
Primary Storage
with Local
Snapshots
Primary Storage
with Replicated
Snapshots
Availability
Storage
Admin

Integrates primary and protection
storage to:
Eliminate impact on application
Gain faster backup and recovery
Reduce cost and complexity
INTRODUCING EMC PROTECTPOINT
Solving the gap in todays protection solutions

EMC PROTECTPOINT BENEFITS
Eliminate Backup
Impact on App
Meet Stringent
Protection SLAs
Reduce Cost and
Complexity
Non-intrusive
Data Protection
Consistent Application
Performance
Faster Backup
and Recovery
Meet SLAs for Mission
Critical Applications
Simple and
Efficient
No Additional
Infrastructure Required
Reduce Time, Cost & Complexity in Managing Backups

RPO = MINUTES
MINIMAL RTO
COST EFFECTIVE
RELIABLE RECOVERY
APP INTEGRATION
BACKUPS SNAPSHOTS+
The functionality of backup with
the performance of snapshots
BEST OF BOTH WORLDS DATA PROTECTION
NO APP IMPACT

Data Domain
system
App Server
MANAGE
Primary Storage
App OwnerStorage Admin Backup Admin
SETUP MONITOR
Data Domain
system
CONFIGURE ANALYZE REPORT
Enabling all persona to do their jobs through their own tools
MANAGEMENT PLANE - PROTECTPOINT
• Storage Admin
– DDSM
– DDMC
– Unisphere
• App Admin
– ProtectPoint
Agent
• Backup Admin
– DPA

• At ProtectPoint DA time File System based Oracle backup is supported
• Technically any database backup can be supported by
– Writing custom scripts
– Engaging with EMC Professional Services
– Implementations in virtual environments like VMware with RDM or
Microsoft Hyper-V can be supported via RPQ
– Oracle RAC and ASM may be considered under RPQ
• VMAX3
• DD4500, DD7200, DD990 with DD OS 5.5.1
SUPPORTED WORKLOADS & CONFIGURATIONS

NEW TECHNOLOGY POWERING PROTECTPOINT
• VMAX3
– Change Block Tracking
• Data Domain
– Block Storage Services
Full Backup
Fiber
Channel
Block
1. Dedupe
2. Compress
3. Write to File System
New Block
Production
New Block
Point in
Time Copy
ProtectPoint Agent on
the Application Host
More than just layering software on top of snapshot

• Data Domain has new Block Services (FC/FCP)
– Adding support to be an FC target for the VMAX3
– Ability to consume the change blocks form the VMAX and create
independent full deduplicated copies in native format
– Managed File Replication
• Supported only on DD4500, DD7200, DD990 for ProtectPoint
– DD OS 5.5.1 or higher
HOW DO WE DO THIS - DATADOMAIN

• Data Domain Abstraction through FTS
– Allows external SCSI devices to be treated as internal VMAX
device
– On all new VMAX but only supported for ProtectPoint at GA
• New Change Block Tracking Mechanism
– More space efficient and higher performance and scale than prior
technology
– Designed to enable ProtectPoint as well as snapshots
HOW DO WE DO THIS - VMAX

Application
Operating
System
ProtectPoint Agent
Application Server (Host)
Storage
Configuration
Credentials
RSA Lock Box
Backup Catalog
Application Layer
(File System CLI)
ProtectPoint
Controller
VMAX
Devices Backups
PROTECTPOINT CONTROLPLANE & HOSTAGENT

Data Domain
system
App Server
MANAGE
Primary Storage
App OwnerStorage Admin Backup Admin
SETUP MONITOR
Data Domain
system
CONFIGURE ANALYZE REPORT
Enabling all persona to do their jobs through their own tools
MANAGEMENT PLANE - PROTECTPOINT
• Storage Admin
– DDSM
– DDMC
– Unisphere
• App Admin
– ProtectPoint
Agent
• Backup Admin
– DPA

Backup
Images
Unique Blocks
Ingested
Production
Data
Point In
Time Copy
CBT Transfer
Deduplication IndexChange Block Tracking Engine
Inline
Dedupe
DURING SETUP – GENERATION ZERO BACKUP
Storage Admin
SETUP
CONFIGURE
This causes the
initial transfer of
data to the Data
Domain but does
not create a
backup
Thus even first
backup only sends
change traffic

App
OS
Backup
Images
Unique Blocks
Ingested
Production
Data
Point In
Time Copy
SAN CBT Transfer
Deduplication IndexChange Block Tracking Engine
Inline
Dedupe
DATA PLANE - BACKUP WITH PROTECTPOINT
App Owner

App
OS
Backup
Images
Unique Blocks
Ingested
Production
Data
Point In
Time Copy
SAN
Dedupe
Index
Recovery
Device
Granular Recovery via Instant Access
Recovery
Device
Mapping
RESTORE WITH PROTECTPOINT
App Owner For more detail, consult: ProtectPoint White Paper

EMC Vmax3 tech-deck deep dive

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to EMC Vmax3 tech-deck deep dive (20)

More from solarisyougood (20)

Recently uploaded (20)

EMC Vmax3 tech-deck deep dive