3. Proxy Server
Site B
(Physical)
Proxy Server
Site A
(Physical)
Veeam ONE
WAN
Accelerator
Site B
VMFS
Datastore
WAN
Accelerator
Site A
Veeam ONE
(Optional)
WAN
Accelerator
Site A
Proxy Server
Site A
(Physical)
Proxy Server
Site B
(Physical)
WAN
Accelerator
Site B
VMFS
Datastore
Disaster Recovery
Replication Topology
Veeam Proxy
WAN Accelerator
Veeam ONE
Veeam BR
Virtual Appliance Proxy
Replication Datastore
WAN
Storage
Connections
(Direct SAN)
External
Network
Internal
Network
Replication
Internal
Network
4. Disaster Recovery
Backup Infrastructure Overview
Role: Veeam Backup & Replication Server
Role Type: Management Server
OS Type: Windows Server x64 – 2008 R2 And Above
Machine Type: Virtual
Per site: 1 Server Per Site
CPU: 4 Cores
Memory: Minimum 4GB – 500MB Per Each Concurrent Job
Disk: Depended On SQL DB And Meta Data Size – Minimum 80GB
Database: Local SQL Server
Network: 1 x vNIC – 1Gb
Role: Veeam ONE Server
Role Type: Monitoring & Reporting Server
OS Type: Windows Server x64 – 2008 R2 And Above
Machine Type: Virtual
Per site: 1 Server Per Site
CPU: 4 Cores
Memory: Depended On Virtual Environment Size – Minimum 8GB
Disk: Depended On SQL DB Size – Minimum 80GB
Database: Local SQL Server
Network: 1 x vNIC – 1Gb
Role: Veeam WAN Accelerator
Role Type: Cache Server
OS Type: Windows Server x64 – 2008 R2 And Above
Machine Type: Physical
Per site: Recommended Two Per Site
CPU: 8 Cores
Memory: Minimum 12 GB
Disk: Calculated On Next Slides – Local / Remote
Database: No DB
Internal Network: 2 x NIC – 10 Gbps
External Network: 1Mbps And Faster WAN Connectivity
Role: Veeam Proxy
Role Type: Transport Machine
OS Type: Windows x64 – Windows 7 And Above
Machine Type: Physical And Virtual
Per site: Recommended Two Machines Per Site
CPU: 1 Core Per Each Concurrent Task
Memory: 200MB Per Each Concurrent Task
Disk: Physical Connectivity To SAN Fabrics*
Database: No DB
Network: 2 x NIC – 10 Gbps
*Physical connectivity will be used by physical proxy, virtual appliance will use hot-add feature.
5. Disaster Recovery
Requirements – Replica Storage Calculation
Thick VMDK
100GB
Replica With Thick Disk
Memory Size
8GB
Change Rate
(Incremental)
*
Retention
1GB * 4
Thick Replica
Size
112 GB
+
50% Working
Space
Thick VMDK
100GB
Average
Utilization
80GB
Replica With Thin Disk
Memory Size
8GB
Change Rate
(Incremental)
*
Retention
1GB * 4
Thin Replica
Size
92 GB
+
50% Working
Space
• Change rate will be calculated during a period of time and we can retrieve it from Veeam ONE Reporter.
• Working space should be guaranteed, because of future replication process and snapshot creation.
6. Disaster Recovery
Requirements – WAN Accelerator Storage Calculation
20GB
Free Space
Per Each 1TB
VMDK
Source WAN Accelerator
10% Guard
22GB
Total
100GB
Minimum
Recommendation
10GB
Free Space
Per Each OS
Type
Target WAN Accelerator
10% Guard
33GB
Total
100GB
Minimum
Recommendation
• Global cache is stored only on the target WAN accelerator. We do not have to provide space for global cache on the source WAN accelerator.
20GB
Free Space
Per Each 1TB
VMDK
7. Disaster Recovery
Consideration – Storage Performance
Four Active Snapshot Per Datastore
• Preventing impact on production storage performance and affect virtual machines performance. Can be
increased according to storage device load.
Dedicated Storage
• Having dedicated storage space on a storage device or dedicated LUNs to preventing impact on recovery site
storage performance and affect virtual machines performance.
• Bottleneck is always source during data processing, so use faster disks will help to reduce impact on production
virtual machines.
Direct SAN Access
• Offload on first session replication by accessing to storage device directly via HBA and SAN fabric.
• Data changes will be transferred via other available transport mode ( NBD or Hot-Add).
• Can be used for Thick disk restoration only, otherwise disk should be converted.
Virtual Appliance Proxy
• All LUNs must be accessible via ESXi storage connections.
• Adding more SCSI Controller, increasing concurrent replication tasks.
8. Disaster Recovery
Consideration – Replication Time
Average Rate
Of Change
Data Size
20GB
WAN
Bandwidth
10 Mbps
Bandwidth
Usage
90%
Replication
Time
300 Minutes
• Bandwidth calculation formula is: ((Avg. Of Change Rate) / (90% Of Bandwidth MBps)) / 60 Seconds
• Replication time will be increased based on data processing time and storage performance
WAN Replication Time
Data Change Processing Time
Need To Measurement At Real Situation
9. Disaster Recovery
Consideration – Network
Internal Network Bandwidth
• Network transport mode (NBD) has good performance on 10Gbps and faster links.
• 2 x 10Gbps links should be available for each Veeam proxy.
Isolated Internal Network
• Dedicated network switch for backup and replication to replication traffic isolation.
• Dedicated virtual switch on ESXi servers to replication traffic isolation.
Network Accessibility
• All replication infrastructure components should be accessible on internal and external network, from
production site or recovery site.
• All required ports should be opened on firewall and allow in ACLs, otherwise backup and replication jobs will be
failed.
External Network Bandwidth
• Continuously processing data and replicating data between production site and recovery site needs fast WAN
link with low lose packet. Minimum 1Mbps bandwidth is needed.
10. Disaster Recovery
Consideration – Infrastructure
Veeam Proxy
• Veeam proxies can be shared between Veeam backup servers, so each backup server needs one source proxy in
source side and one target proxy on target side.
Veeam WAN Accelerator
• Veeam WAN Accelerators can be shared between Veeam backup servers, each backup servers needs one WAN
source and WAN target at least.
Domain And Credential
• All Veeam replication components should be joined or disjoined to one domain. If vCenters are joined to
different domains, local credentials are preferred for connection.
NBD Limitation
• Regarding to limitation on vSphere API to establishing 7 concurrent network connections to each ESXi server,
Veeam proxies can’t processing more than 7 jobs by default.
11. Disaster Recovery
Replication / Failover / Failback Process
Communicate
Between Data
Movers In
Both Side
Copy VM Data
In Source Side
By Veeam
Proxy
Filtering
Overlapping,
Zero Data And
Swap Data
Blocks
Check
Metadata To
Detect Block
Changed
Copying Data
Blocks,
Compressed
And Move To
Target Side
Decompress
Replica Data
And Write To
Target
Datastore
Continuously Replication
Starting To
Write
Changes To
Delta
Snapshot
Do Network
Re-Mapping
And Run Re-IP
Rule
Power On
Replica VM –
Change Status
To Normal
Roll Back VM
Replica To
Required
Restore Point
Failover Process
Check
Difference Again
And Send
Changes To
Original VM
Again
Power Off
Replica VM
Transport
Changed Data
To Original VM
Check
Difference
Between Replica
And Original
VMs
Create Failback
Snapshot From
Original VM
Power Off
Original VM
Failback Process
12. Disaster Recovery
Next Step
Before Production Implementation
Simulating replication at real situation with some production servers
Storage space measurement
Network bandwidth measurement
Checking limitation
Detecting problems on real situation