BUILDING THE VIRTUAL DATA CENTER THROUGH

AVAILABILITY, MOBILITY, COLLABORATION

Ankur Patel Sr. Systems Engineer

2

• Why VPLEX?

• DR vs HA

• Current deployments

• Hardware/Software Overview

• Use cases

• Don’t forget about DR

• Distributed Cache Coherency

• Implementation

• Summary / Q&A

Agenda

3

• ‘Building Blocks’ of resources

• Load Balancing across sites

• Federated data repository

• Continuous or High Availability of applications and data

• Optimal routing and response times

The Distributed Data Center

4

Key: ● Good Solution; ● May Provide Some Coverage; ● No Coverage

Availability Failure/Coverage Matrix Failure

DR Outage<RTO

HA Same Site

CA - 2-Site HA

Scheduled Events – ~85% of Outages Site Maintenance (whole site or component) ● ● ● Software – application upgrade/same DB ● ● ● Software – application upgrade w/DB upgrade ● ● ● Software – Agents, BIOS, OS, & Utility ● ● ● Unscheduled Events – ~15% of Outages

Hardware Component Failures ● ● ● Human Process Error ● ● ● Network ● ● ● Power ● ● ● SAN Failure ● ● ● Site Outage - Recoverable ● ● ● Disaster Events: < 1% of Outages

Site Loss – Catastrophic (outage > RTO) ● ● ● Very Large Scale Regional Disaster (outage > RTO) ● ● ●

5

• 220+ PB DEPLOYED WITH ACTIVE/ACTIVE • 2000 clusters deployed • Largest deployment to date – 16PB

VPLEX Momentum

OVER 6-9’s SYSTEM UPTIME

OVER 11 MILLION RUN HOURS

OVER 40 SUPPORTED PLATFORMS

“A unique (and importantly, integrated) approach to addressing the persistent problem of integrating distributed data centers and making the (IT) world smaller.” - IDC

“A ‘next-generation’ wake-up call…VPLEX constitutes a bold vision – a game changer.” - The Clipper Group

“An innovation with the potential to change the way organizations deploy and manage storage infrastructure.” - Enterprise Storage Group

6

EMC VPLEX with GeoSynchrony • What is it?

– Virtual storage platform – Located in the SAN – Federated heterogeneous

storage • How is it unique?

– Distributed coherent cache – Fully redundant scale-out

clusters – Distributed storage cluster with

arbitration witness

AccessAnywhere

The VPLEX Family

VPLEX Local

Single Cluster

VPLEX Metro

Synchronous

Distributed. Dynamic. Smart.

VPLEX Geo

Asynchronous

VPLEX Global

Synchronous or Asynchronous

VISION

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.

8

VPLEX– Hardware

• Non-disruptive hardware/software upgrades • 8000 LUNs / 1600 initiators/ 400 IT Nexuses per FE • Engine max IOPS ~50k-80k OLTP2 real world

9

Rack Configuration

10

VPLEX BE Connectivity

11

VPLEX FE Connectivity

12

Leveraging VPLEX

• Availability

• Migration

• Mobility

Information Where It’s Needed, When It’s Needed

Total Infrastructure Resiliency

• Continuous data availability for mission-critical applications. • Sustain failure of any component within the infrastructure without the

need to failover host applications between sites. • Data R/W accessible at both sites under normal operation.

Failure situation

Action

Server VMware HA automatically restarts affected VMs on other side

VPLEX Cluster VPLEX cluster witness support enables surviving side to continue processing I/O VMware uses path failover to direct I/O to remaining VPLEX

Storage VPLEX automatically redirects I/O to surviving storage

ISL If Active/Active config/ may require scripted restart in order to avoid conflicting detach If Active/Passive – active side continues

14

Availability: VPLEX Witness Cluster Arbiter

• Coordinates seamless failover

• Enables longer distance for stretched server clusters

• Provides 0 RTO for storage changes or events

VPLEX WITNESS

• < 1 second response time

• Third site, virtual machine

Non-Disruptive Migrations and Workload Rebalancing

Continuous Operations With VPLEX

Removes the application downtime associated with storage array refreshes

Enables storage optimization by enabling restructuring of data across storage arrays

Data center migrations can be done as an on-line event

AOL had migrated 48 arrays in 52 weeks from May 2010 - May 2011 – an average of an array every 5.4 business days

Data Mobility

Dev A

Vendor A Vendor B

Zero impact to host Applications when moving data

Active Failure Domain

Vendor C

Active Failure Domain

Data Mobility

Dev C

Dev B

Storage Migration

Optimization

Datacenter Migration

Rebalancing

LUN 1v

LUN 1v

Distributed Virtual Volume

Use Case

Contemporary Solution

Availability: VPLEX Witness + Cross Connect

• Ideal for shorter distances, response time <1ms

• Leverages multipathing SW

• Trade off of higher host availability and CPU cycles

• Reduces server failovers

AccessAnywhere

VPLEX WITNESS

Active Active

Site 1 Site 2

Site 3

VPLEX Local Use Cases MOBILITY

Simplify recurring data movement

Heterogeneous nondisruptive technology refreshes

Move data without impacting users to reduce downtime

MOBILITY

SIMPLIFIED PROVISIONING

Aggregate capacity and improve utilization

Standardize LUN presentation and management

AVAILABILITY

Sustain component or array failures without host impact

Increase availability during failures

Before VPLEX: Typical Config.: Oracle RAC with Array-Based Replication for Disaster Recovery

ACTIVE PASSIVE

Array-Based Remote

Replication

Areas for Enhancement

Requires full failover and restart at remote site

Recovery Time Objective (RTO) is

long

Complex DR Testing

RAC With ASM Mirroring for Extended Distance

LAN Backbone

RAC interconnect

LAN Backbone

Mirror A (active) Mirror B (active)

SAN inter-switch link

Mirrored I/O paths

RAC Quorum

Disk

Host Mirroring Consumes Valuable CPU Cycles

Cross-Connect Introduces SAN Complexity

Admins need to configure failure groups, path preferences and generic ASM mirroring functions

As storage capacity and node counts increase, all of the above need to be re-administered for balanced performance and HA requirements

Quorum Disk Management For Remote Replication

Site 1 Site 2

LAN Backbone

RAC interconnect

LAN Backbone

Mirror A (active) Mirror B (active)

VPLEX interconnect

VPLEX Witness

“Identical” virtual volumes (actually the same volumes)

Site 1 Site 2

RAC With VPLEX Metro for Extended Distance

No host mirroring which saves valuable cpu cycles

No need to merge fabrics which reduces SAN complexity and improves fault domain isolation

No need for Oracle DBA to configure failure groups, path preferences and generic ASM mirroring functions

As storage capacity and node counts increase, VPLEX handles path balancing, scaling storage, non-disruptive movement of data and replication workload

No 3rd site quorum disk administration as quorum devices reside on VPLEX

Mobility: Where can you deploy?

Metro Geo

(Live)

(Live)

(Live) OVM

(Live)

LPAR (Live)

VPLEX Front-End Conceptual Design Wide IP

BACK-END FEDERATION

FRONT-END ACCESS

Utilize Existing Replication

VPLEX

Brocade, Cisco

VMAX

Brocade, Cisco

DMX

SRDF/A

DMX

Brocade, Cisco

VPLEX

2

EMC RecoverPoint Overview

SAN SAN

Continuous Data Protection (CDP)

3rd

Party

Application servers

File and Print servers

Database servers

WAN SAN

Continuous Remote Replication (CRR)

3rd

Party

Application servers

File and Print servers

Database servers

Remote site disaster recovery

Roll back of replicated data (“Near” CDP)

Target Side Processing (TSP) of replicated data

Policy-based lag and WAN compression

Local, transaction-level data protection

“True” CDP (Any-Point-in-Time) recovery

Consistency groups with bookmarks

RecoverPoint CDP RecoverPoint CRR

EMC RecoverPoint Overview

Leverage RecoverPoint For Third Site Protection and CDP

Replication: VPLEX + RecoverPoint

NetApp FAS, 3xxx, 6xxx (Block)

IBM DS, XIV, SVC

EMC VNX, VMAX, HP 3PAR, etc.

RecoverPoint CRR

RecoverPoint CDP

VPLEX Metro

IBM DS, XIV, SVC

NetApp FAS, 3xxx, 6xxx (Block)

26

Cache

Cache Directory D

Cache Directory F

Cache Directory H Cache Directory B

Distributed Cache Coherency

Cache Directory C

Cache Directory E

Cache Directory G

Cache

Engine Cache Coherency Directory Block Address 1 2 3 4 5 6 7 8 9 10 11 12 13 …

Cache A

Cache C

Cache E

Cache G

Engine Cache Coherency Directory Block Address 1 2 3 4 5 6 7 8 9 10 11 12 13 …

Cache A

Cache C

Cache E

Cache G

Cache Directory A

New Write: Block 3

Read: Block 3

Cache Cache

Directory-based distributed cache coherence efficiently maintains cache state consistency across all engines

27

1) Server is shutdown

Servers

Donor Array

VPLEX

1

Encapsulation Method (Short outage to cutover)

28

Encapsulation Method (Short outage to cutover)

1) Server is shutdown

2) VPLEX “Claims” the volumes that were presented to the server

3) Claimed Volumes are configured as Devices on the VPLEX and presented back to the host.

Servers

Donor Array

VPLEX

1

2

3

29

Host Migration Method (On-line - No outage to cutover)

1) New VPLEX Target Volumes are presented to the Server via the SAN

2) Host Migration to VPLEX • VMware Storage Vmotion

• AIX LVM Mirroring

• Solaris/ Veritas LVM Mirroring

3) Once Migration is complete, the Donor Mirror can be broken and the migration to the VPLEX is complete

Servers

Donor Array

VPLEX

1

2

Summary Benefits

Business Benefits - VPLEX Benefit Area(s) Impact Customer Value

Reduction in Planned Downtime (non disruptive data and/or virtual machine mobility)

• DR testing/DC Flip-Flop • Power Outages • Storage maintenance or refresh • Fabric maintenance or refresh • ESX Server maintenance

Increase in Business Application Availability

Lost revenue or customer satisfaction down to application not being available.

Reduction in Unplanned Downtime

• Active/Passive application cluster (e.g. VMware HA)

• Automated failover • Reduced failover time • Simplification

• Active/Active application cluster (e.g. Oracle RAC, Sanbolic AppCluster for SQL)

• Zero downtime even with the loss of a datacenter

Increase in Business Application Availability

Reduce lost revenue or customer satisfaction down to application not being available.

Asset Utilization • Storage pooling • Sweat existing assets

• Use Storage and Compute Assets in secondary DC for Production I/O

Increased application performance due to distributed workload and VPLEX cache capability

Delayed or no additional expenditure due to increased I/O requirements

Operational Benefits - VPLEX Benefit Area(s) Impact Simplified Storage Management

In a multi-vendor or multi-model storage environment where all of the back-end storage has been presented to VPLEX, having a single management interface for all storage related provisioning tasks could reduce complexity in the environment

Reduce complexity

Common Host Build Deploying VPLEX creates an abstraction layer between the hosts and the physical storage, with hosts only seeing VPLEX as the storage array. This means that only one host build is required to support VPLEX, irrespective of the underlying storage hardware being virtualized

Reduce complexity and cost

Simplified Application Recovery

Application restart in the event of infrastructure or site failure can be complex and time consuming. Using VPLEX Metro configurations in conjunction with Operating System and Application Clustering technologies enables organizations to create ‘Local’ Clusters across geographic distances.

Reduced complexity, HA for applications that do not have a geographic clustering capability

Standardize on replication technologies

Replication between primary datacenters and heterogeneous storage arrays is performed by VPLEX simplifying replication requirements and potentially reducing the cost of replication licenses and ongoing software maintenance.

Reduce complexity

Non-disruptive data mobility

Data can be moved between storage arrays non-disruptively, for example in case of a storage refresh or storage tiering across across arrays without the need for application outages.

Reduce complexity and enable tasks to be conducted during normal working hours

Local Heterogeneous Mirroring

When mirroring across disk arrays using host based logical volume managers, creating, maintaining and resolving mirrored disk issues is performed on a host-by-host basis. VPLEX replaces the need to mirror at the host layer.

Reduced cost and complexity