1EMC CONFIDENTIAL—INTERNAL USE ONLY
EMC PROTECTION FOR MICROSOFT EXCHANGE SERVER 2010EMC VNX5700, EMC RecoverPoint, EMC Replication Manager, EMC Data Protection Advisor, VMware vSphere, and VMware vCenter Site Recovery Manager
• Fully automated recovery for Microsoft Exchange Server 2010• Unlimited snapshots to protect from logical database corruption• In-depth monitoring and reporting
EMC Solutions Group — Strategic Solutions Engineering
2EMC CONFIDENTIAL—INTERNAL USE ONLY
Solution Offering and Options• Full end-to-end Protection for Exchange 2010 environment in a
virtualized infrastructure– Local application protection with Exchange deployed in DAG– Local continuous data protection (CDP) with RecoverPoint
(unlimited snaps)– Remote continuous replication (CRR) with RecoverPoint for
DR– Automated failover and failback with vCenter SRM 5
• Database protection and instant restore with Replication Manager
• In-depth monitoring and reporting with DPA• Flexible deployment options• White paper for this solution is published here:
http://www.emc.com/collateral/hardware/white-papers/h8891-emc-protection-fr-microsoft-exchng-servr2010.pdf
3EMC CONFIDENTIAL—INTERNAL USE ONLY
Exchange 2010 DR OptionsExchange 2010
DR OptionsReplication
Method Description Reference Material
Cross-site (stretch) DAG
Exchange Continuous Replication (DAG)
Built in to Exchange 2010 for HA and DR
Co-branded White Paper: Business Continuity for Microsoft Exchange 2010 Enabled by EMC Unified Storage, Cisco UCS and Microsoft Hyper-V
Database Portability
RecoverPoint Just the Exchange data is replicated. The users are re-homed at the second site to a different server
White Paper: Exchange 2010 Disaster Recovery Options with Cross-Site Exchange DAG and EMC RecoverPoint
Replication Enabler for Exchange (REE)
RecoverPoint Synchronous solution today. Integrated with Exchange 2010 3rd Party Replication API. Replaces native DAG replication
Co-branded White Paper: Zero Data Loss Disaster Recovery for Microsoft Exchange 2010 Enabled by EMC Unified Storage, EMC Replication Enabler for Exchange 2010 and Microsoft Hyper-V
Server/Site Move
SRDF, RecoverPoint
Both the OS and Exchange data is replicated. Failover includes bringing the servers up, changing IP & updating DNS (automated with SRM 5)
White Paper – SRDF Solution: Virtualized Exchange 2010 DR Enabled by EMC Symmetrix VMAX and VMware vCenter SRMWhite Paper – RecoverPoint Solution: EMC Protection For Microsoft Exchange Server 2010
4EMC CONFIDENTIAL—INTERNAL USE ONLY
Solution Architecture• Two active-active Sites• 10,000 Exchange 2010 users per site (20,000 total)• 1 DAG per site with 2 local copies• 4 Mailbox Server VMs per DAG (each VM configured to run
5,000 users during local switchovers — 2,500 active/2,500 passive)
• Exchange user profile• 2 GB mailbox per user• 0.15 IOPS per user
5EMC CONFIDENTIAL—INTERNAL USE ONLY
Data
Site Protection with RecoverPoint and Automated Failover with SRM
RecoverPoint
RecoverPoint
Site A Site B
WAN Link
Data
SRM
6EMC CONFIDENTIAL—INTERNAL USE ONLY
Parameters Used to Test and Validate the SolutionParameters Local DAG with RecoverPoint CRR environmentNumber of sites Two active-activeNumber of users per site 10,000 users (20,000 total during site failover)Exchange 2010 user profile 150 messages sent/received/day (0.15 IOPS), 2 GB mailboxDatabase read/write ratio 3:2 in Mailbox Resiliency configuration
Test Environment • 4 mailbox servers per DAG (local site)• 3 copies (2 local with DAG and 1 remote with RecoverPoint)
Connectivity between sites 1 Gbps Ethernet Storage connectivity to physical host FC
Storage design• 1 VNX5700 storage array per site• 2 TB NL-SAS (7.2k rpm) disks for Exchange database and logs (RDM/P)• 2 TB NL-SAS (7.2k rpm) disks for virtual machine OS (VMFS)• 600 GB SAS (10k rpm) disks for RecoverPoint journals
vSphere host design
• 4 Intel-based servers with 4 six-core Xeon X7460 processors @ 2.66 GHz (24 CPUs) and 128 GB RAM. 2 servers per site, each supports 4 Exchange Mailbox Role virtual machines.
• 4 Intel based servers with 4 six-core Xeon X7350 Processors @ 2.93 GHz (16 CPUs) and 64 GB RAM. 2 servers per site, each supports 4 Exchange Client Access and Hub Transport Role virtual machines.
Virtualization Platform vSphere 5.0
Data protection• Daily full backups (hardware VSS)• 24 application-consistent snapshots per day (24 hours)• Unlimited crash-consistent snapshots per day (24 hours)
7EMC CONFIDENTIAL—INTERNAL USE ONLY
ESXi Host & VM Design
Exchange 2010 Role CPU MemoryMailbox Server 6 32Client Access and Hub Transport Combo 4 16
NOTE: CPU and memory requirements are based on Microsoft Guidelines. See http://technet.microsoft.com/en-us/library/ee712771.aspx for additional details.
8EMC CONFIDENTIAL—INTERNAL USE ONLY
Mailbox Server Role Design
9EMC CONFIDENTIAL—INTERNAL USE ONLY
Exchange 2010 Mailbox Server VM DesignESX
Server 1 ESX
Server 2 MBX1 MBX2 MBX3 MBX4DB1 C1 C2 DB2 C1 C2 DB3 C1 C2 DB4 C1 C2 DB5 C1 C2 DB6 C2 C1 DB7 C2 C1 DB8 C2 C1 DB9 C2 C1 DB10 C2 C1 DB11 C1 C2DB12 C1 C2DB13 C1 C2DB14 C1 C2DB15 C1 C2DB16 C2 C1DB17 C2 C1DB18 C2 C1DB19 C2 C1DB20 C2 C1
Design Attributes• 5,000 users during local switchovers
(2500 active/2500 passive)• 10 DBs per server• 500 users per DB• 1.6 TB DB LUNs• 100 GB Log LUNs• VNX pools• Physical isolation of DAG copies
Active Copies Passive Copies
Domains protected against failure• Mailbox server• Hypervisor host• Storage• Network
10EMC CONFIDENTIAL—INTERNAL USE ONLY
Mailbox Server Building Block Design
Users per Exchange Mailbox server virtual machine
Disks per Exchange Mailbox server virtual machine
vCPUs per Exchange Mailbox server virtual machine
Memory per Exchange Mailbox server virtual machine
5,000 20 (16 + 4):• 16 x 2 TB NL-SAS disks
in RAID 1/0 for databases
• 4 x 2 TB NL-SAS disks in RAID 1/0 for logs
6 CPUs 32 GB
11EMC CONFIDENTIAL—INTERNAL USE ONLY
Storage Design Details for a 5,000-User Mailbox Server Building Block
Attribute Details
Number of users 5,000 (in a switchover condition)
User profile 150 messages sent/received (0.15 IOPS), 2 GB Mailbox
Disk type 2 TB NL-SAS, 7.2k rpm
Storage configuration/RAID type 20 disks per building block for Exchange data:• 16 disks in a RAID 1/0 storage pool for databases• 4 disks in a RAID 1/0 RAID group for logs
Databases per server/users per database 10 databases per server, 500 users per database
LUNs/LUN sizes 21 LUNs per Mailbox server virtual machine:• 10 database LUNs (1.6 TB) (RDM/P)• 10 log LUNs (100 GB) (RDM/P)• 1 LUN for virtual machine OS (200 GB) (VMFS)
12EMC CONFIDENTIAL—INTERNAL USE ONLY
ESX/VMs Storage Design
13EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint DesignIn this solution, EMC RecoverPoint provides protection from Exchange logical database corruption and DR protection in the case of a site failure.
• Native Exchange replication does not protect from logical corruption.
• Corruption can potentially replicate to other Exchange database copies within the DAG locally and remotely.
In this solution, RecoverPoint is configured as follows:
• The replication method is Concurrent Local and Remote Replication (CLR) (combines both CDP and CRR).
• Each Mailbox server VM is placed into its own CG with all replicated database and log LUNs, including the LUN containing the guest OS.
• Ensures full consistency across all volumes used by the Mailbox Server Role virtual machines.
14EMC CONFIDENTIAL—INTERNAL USE ONLY
Consistency Group (CG) Design for Active/Active Data Centers
• Ensures consistency across production and targets – one Exchange Mailbox server per CG
• Application recovery can be independent from SLAs
• Enables independent replication of various applications
• Utilizes Consistency Groups to organize data
MBX3 CG
MBX1 CGCDP
CRRCDP MBX2 CG
MBX1
MBX2
CRR
CRRMBX3 CDP
MBX4 CG CRRMBX4 CDP
CDP
CDP
MBX1
MBX2
MBX3 CDP
MBX4 CDP
CRR
CRR
CRR
CRR
MBX5 CG
MBX6 CG
MBX7 CG
MBX8 CG
Site A Site B
15EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Consistency Group Configuration One Consistency Group per Mailbox Server VM 21 Replication Sets per Consistency Group
16EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Group SetsA Group Set is a set of consistency groups to which the system applies parallel bookmarks at a user-defined frequency. Allows for automatic, periodic bookmarking Provides same RPO for multiple Consistency Groups distributed across multiple RPAs Useful for consistency groups that depend on each other or must work together as a
single unit• Four Mailbox servers in a DAG at each site – combined into one Group Set to meet 15-
minute RPO requirement
17EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint CG Policy Settings for SRM
18EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Journal Sizing• RecoverPoint journal volumes store consistent point-in-time
snapshots (or bookmarks) to enable point-in-time recovery of Exchange data.
• Journal size will dictate how much protection you will have (how many snaps you can keep).
• For each CG in the RecoverPoint CLR configuration, three journals are set up:• Two at the local site• One at the remote site
• The two most important considerations when sizing storage for journal volumes: • Performance• Protection window
19EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Journal Sizing (cont.)Formula to calculate journal volume size based on the expected change rate and required rollback time period:
Example• Protection requirements: 3 days (72 hours)• Change rate: 5 Mbps
Journal size = 5 Mbps x 259,200 seconds / 0.8 x 1.05 = 1,701,000 Mb (~213 GB)
20EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Journal Sizing (cont.) Journals deployed in this solution
21EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Storage Design Summary
Volume type Site A and B CommentsRepository 3 GB One volume per storage array/RPA cluster per siteJournal 40 x 100 GB
32 x 5 GB• Journal LUNs configured from RAID 1/0 groups.
• 500 GB journal per consistency group: 5 x 100 GB volumes for each local and remote copy.
• 20 GB journal per consistency group: 4 x 5 GB volumes for each production copy.
Replication 80 x 1.6 TB for database volumes80 x 100 GB for log volumes
• RAID 1/0 pools• Replication volume sizes must be equal or larger
than production volumes
22EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Integration with Replication Manager RM leverages RP replication technology to automate and
schedule VSS copies of Exchange, both locally and remotely RM automates Exchange 2010 tasks such as:
• Create Exchange database application sets• Create and manage multiple-site, application-consistent replicas of
Exchange databases• Run a full Exchange job to create a VSS bookmark, run the Microsoft
Exchange utility eseutil, and then truncate the Exchange logs• Run an Exchange copy job that creates a VSS bookmark• Start an on-demand mount and dismount of an Exchange database• Restore Exchange databases from either local or remote bookmarks
to the production Exchange server in seconds
23EMC CONFIDENTIAL—INTERNAL USE ONLY
How this solution uses RMRM automates RecoverPoint application-consistent replicas and performs Exchange log truncation 24 application-consistent (VSS) snapshots in a 24-hour cycle
• In addition to maintaining hourly application-consistent snapshots initiated by RM, RP journal maintains many more crash-consistent snapshots from which Exchange databases can easily be recovered
RM application set is created for each Exchange Mailbox server virtual machine.
Application set includes all ten databases configured on the server• This design aligns with the RP CG design, where each consistency
group contains all of the Exchange databases and logs for a specific Mailbox server
24EMC CONFIDENTIAL—INTERNAL USE ONLY
Replication Manager Design RM Application Set
configuration– Created for each Mailbox VM– Includes 10 DB’s per
Mailbox Server– Design aligned with
RecoverPoint
25EMC CONFIDENTIAL—INTERNAL USE ONLY
Special Environment Variable for RM In this solution each RecoverPoint consistency group is configured
with an additional replication set for the Mailbox server OS LUN that resides on the VMware VMFS datastore volume
• This configuration creates a potential conflict between Replication Manager and RecoverPoint, because the Microsoft VSS framework does not support VMFS data stores
• To resolve this conflict, environment variable is created on each Exchange Mailbox server virtual machine with Replication Manager Agent installed:
ERM_RP_EXTRA_DEVS_ALLOWED=1 With this environment setting in place
• RecoverPoint continues to provide crash-consistent replicas of all LUNs attached to the Exchange Mailbox server virtual machine (including the OS VMFS volume)
• Replication Manager continues to provide application-consistent replicas of all Exchange RDM volumes
26EMC CONFIDENTIAL—INTERNAL USE ONLY
RM Replication options In this solution, RM jobs are configured with two Replica Type
options to protect the Exchange Server 2010 databases:– Online Full– Online Copy
Online Full jobs are set up for daily backups Online Copy jobs are set up to run at hourly intervals during
business hours to provide application-consistent recovery (24 snapshots per 24 hours)
27EMC CONFIDENTIAL—INTERNAL USE ONLY
Replication Manager server thaws application
RecoverPoint + RM = Integrated Local/Remote Protection
Local CDP Copy
Exchange Production
Exchange Mount
Host
Local CDP Journal
Production LUN
Replication Manager Server
Replication Manager server freezes application Replication Manager server requests VSS-compliant bookmark
RemoteCRRJournal
WANRecoverPoint RecoverPoint
SANSAN
RemoteCRRCopy
LUNLUN
BOOKMARK
BOOKMARK
28EMC CONFIDENTIAL—INTERNAL USE ONLY
Site Recovery Manager Core Capabilities
Virtual Center
Virtual Machines
ESXServer
ESXServer
ESXServer
ESXServer
Servers
Storage
Site Recovery Manager
Centralized management for DR• Create, test, update and execute recovery
plans from a single point of management• Seamless integration with Virtual Center
Disaster recovery automation• Build recovery process in advance• Automate testing of recovery plans• Automate execution of recovery process
Simplified setup and integration• Allocate and manage recovery resources• Easy integration with leading vendors’
storage replication systems
29© Copyright 2010 EMC Corporation. All rights reserved.
Key SRM ComponentsProtected site
VC Server 1
SRM Server 1
StorageReplicationAdapter
SRM 1 DB
VC 1 DB
RecoverPoint replication
Recovery siteVC Server 2
SRM Server 2
StorageReplicationAdapter
SRM 2 DB
VC 2 DB
RecoverPoint replication
Array 1 Array 2
30EMC CONFIDENTIAL—INTERNAL USE ONLY
vCenter SRM RequirementsSRM requirements for the vSphere configuration at each site:
• Each site must include a vCenter server containing at least one vSphere data center.
• The recovery site must support array‐based replication with the protected site, and the recovery site must have hardware and network resources that can support the same virtual machines and workloads as the protected site.
• At least one virtual machine must be located on a data store that is replicated by RecoverPoint at the protected site.
• The protected and recovery sites must be connected by a reliable IP network. Storage arrays may have additional network requirements.
• The recovery site must have access to the same public and private networks as the protected site, although the recovery site does not necessarily require access to the same range of network addresses.
• VMware tools must be installed on all virtual machines.
31EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Configuration Steps1. Install and configure SRM with the RecoverPoint SRA adapter.2. Configure the connection between the protected and recovery
sites.3. Configure RecoverPoint array managers.4. Configure resource inventory mappings.5. Configure protection groups.6. Create the recovery plan.7. Customize virtual machine recovery options, including IP
customization.
32EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Protection Groups• An SRM protection group is a
collection of virtual machines that all use the same datastore group (the same set of replicated LUNs) and that all fail over together.
• For this solution, an SRM protection group is created for each Exchange Mailbox VM.
33EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Protection Groups (cont.)• After a protection group is created, shadow VMs are automatically
created in the recovery site vCenter inventory.• They act as placeholders for the virtual machines that can be failed
over with SRM.• The shadow VMs cannot be started independently from SRM and are
removed if their protection group is deleted.
34EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Configuration — Recovery Plans
35EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM 5 Features for Multi-subnet Data Centers• Recovery and
Reprotect options
• IP customization
• Power-on priority and dependencies
36EMC CONFIDENTIAL—INTERNAL USE ONLY
Changing the DAG IP address at the Recovery SiteSet-DatabaseAvailabilityGroup -Identity <DAG Name>-DatabaseAvailabilityGroupIPAddresses <new DAG IP-address at recovery site>
37EMC CONFIDENTIAL—INTERNAL USE ONLY
Testing an SRM Recovery Plan• Recovery plans can be tested at any time without
interrupting replication or the business’s RPO.
• Predict how much time the recovery of the virtual environment is likely to take.
• Administrators can quickly access a replica of the entire protected environment for testing purposes.
• Automatically generated summary reports can be used to demonstrate compliance with regulatory requirements.
SRM Recovery Plan Time elapsed (minutes) RTO Met?Single Mailbox server 17 YesFour Mailbox servers in a DAG 25 Yes
38EMC CONFIDENTIAL—INTERNAL USE ONLY
Testing an SRM Recovery Plan (cont.)
39EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Site Failover Test Results
• Failover of all Mailbox Servers from Site A to Site B
• Results include DNS name change and IP changes
SRM recovery plan Time elapsed (min) RTO met?
Four Mailbox servers in a DAG 38 Yes
40EMC CONFIDENTIAL—INTERNAL USE ONLY
Site Failback with SRM• SRM 5 provides automatic failback functionality
• RecoverPoint SRA version 2.0 integrates with SRM 5• This functionality enables you to perform an automated failback
operation after a planned or unplanned failover
41EMC CONFIDENTIAL—INTERNAL USE ONLY
SRM Failback Process1. Verify all network communications between the recovery site and the
original protected site.2. Verify connectivity between vCenter servers.3. Ensure all resource mappings are in place. 4. Ensure no changes need to be made to SRM protection groups or
protected virtual machines.5. Modify the recovery plan to update any custom scripts.
• For example, update Exchange DAG IP addresses that have changed. Also update CAS array IP addresses if the CAS array was previously replicated.
6. To perform failback:• Incrementally—Use the RecoverPoint or Unisphere management console to replicate
data and resynchronize configuration settings from the recovery site back to the originally protected site.
• Automatically—Initiate “Reprotect” from the SRM management interface. 7. When failback process completed, ensure that all Microsoft Exchange
services are started and the databases are mounted.8. Verify email flow and user access.
42EMC CONFIDENTIAL—INTERNAL USE ONLY
Recovery with Replication Manager
43EMC CONFIDENTIAL—INTERNAL USE ONLY
RecoverPoint Journal Usage With Exchange Server 2010 Under Load
Test timeline Journal usage (GB) Journal usage per user (MB)
Journal usage (percentage)*
Start 0.20 ---- 0.054 hrs. (day time) 113 23 MB 30.18 hrs. (day time) 174 36 MB 46.412 hrs. (night time) 184 38 MB 49.016 hrs. (night time) 228 47 MB 60.820 hrs. (day time) 279 57 MB 74.424 hrs. (day time) 345 71 MB 92.0
• LoadGen workload for 5,000 users per server with 150 sent/received messages per user• Hourly application-consistent (VSS) bookmarks for a RecoverPoint consistency group with
all of the LUNs for a single Mailbox server with 10 databases• Automatic crash-consistent (VSS) bookmarks for a RecoverPoint consistency group with
all of the LUNs for a single Mailbox server with 10 databases• Nightly scheduled backup job initiated by Replication Manager
*Percentages are based on a 375 GB usable journal space (500 GB minus 25 percent reserved by RecoverPoint) provisioned for each consistency group and 13.6 TB of replicated Exchange data (80 percent of 17 TB provisioned storage).
44EMC CONFIDENTIAL—INTERNAL USE ONLY
DPA Reporting — RP Journal Utilization
45EMC CONFIDENTIAL—INTERNAL USE ONLY
Summary
Rapid• Automate disaster recovery process• Eliminate complexities of traditional recovery
Reliable• Ensure proper execution of recovery plans• Enable easier, more frequent tests
Manageable• Centrally manage recovery plans• Make plans dynamic to match environment
Affordable• Utilize recovery site infrastructure• Reduce management costs
VMware vCenter Site Recovery Manager and EMC RecoverPoint make Exchange 2010 Disaster Recovery…
White paper for this solution is published here:http://www.emc.com/collateral/hardware/white-papers/h8891-emc-protection-fr-microsoft-exchng-servr2010.pdf