This article documents general best practices and configuration advice when using a Quantum DXi backup storage appliance with deduplication.
Several DXi models can present themselves as a NAS or Veeam Linux Repository appliance for backup purposes. Before you can use a DXi system as a NAS or Veeam Linux Repository appliance, you must first configure a NAS or Veeam Linux Repository share on the DXi.
The default options within Veeam Backup & Replication are intended for non-deduplication storage. As with any modifications to a system that impact performance and/or tuning, your results may vary and are not guaranteed.
Veeam Data Mover Service
The Quantum DXi Linux Repository supports the use of Veeam Data Mover Service, which increases performance between the DXi and the Veeam proxy server by using DXi memory. The Veeam Data Mover performs data processing tasks on behalf of Veeam Backup & Replication, such as retrieving source VM data, performing data deduplication and compression, and storing backed-up data on the target storage. This is important to note as the Veeam Data Mover is used to transfer data rather than NFS, supporting the ability to run the target data mover directly on-board DXi storage, which accelerates certain backup and restore operations and processes.
Veeam Backup & Replication Configuration
Parallel processing (global option):
This option significantly accelerates the backup process and decreases the backup window since virtual disk data is gathered simultaneously. It also dramatically increases fragmentation in the backup files causing high random read for any restore operation. The greater volume of VMs or disks processed simultaneously will increase fragmentation and result in slow restore times. This will increase backup performance, at the cost of reduced restore performance.
Use Per-VM Backup Files (repository option):
Veeam recommends enabling this option so that there is improved performance writing and reading data from the Quantum DXi. This option is enabled by default when adding the repository as a Deduplicating Storage Appliance.
Decompress backup block before storing (repository option):
Veeam strongly recommends enabling this option so that raw data is sent to the Quantum DXi, leveraging its global deduplication and compression capabilities. Leaving Veeam compression enabled may significantly impact Quantum DXi deduplication capabilities resulting in high load and slow backup jobs.
Task limits (repository option):
The DXi Linux Repository supports the use of the Veeam Data Mover Service, which optimizes performance between the DXi and the Veeam proxy server. This optimization uses DXi memory. To avoid oversubscribing memory, we recommend that you run no more than 25 concurrent backups across all repositories defined on the DXi. For recovery jobs, we recommend that you run no more than 15 concurrent full VM restores concurrently across all repositories.
If Quantum DXi is added as Shared folder Repository, this value may need to be adjusted more than once to find the best number of concurrent backups to use for the specific environment. Start low (ie. 5) and increase incrementally to determine when performance starts to fall. There is a direct relation to the proxy maximum concurrent tasks.
Veeam Backup & Replication Job Configuration
Use “Incremental” backup mode. It is not recommended use “Reversed incremental” backup mode.
Check “Create synthetic full backups periodically”.
It is not advisable to perform synthetic full backups when a CIFS Shared Folder repository is employed, please use Active full instead.
Try to limit restore points to from 7 to 14 (Veeam default). In addition, no more than 30 restore points should be retained before a synthetic or full backup is performed
If the decompress setting for the Repository is enabled, this value will not result in compressed data being sent to the DXi. As a result, Veeam recommends leaving this as either AUTO or OPTIMAL for the DXi as it will optimize overall processing.
Inline data deduplication
Disabled, otherwise this option will enable background optimizations for deduplication appliances which affects the performance.
Local target or Local+16TB. Local+16TB produces a performance gain with synthetic full backups (up to 14%). Local+16TB causes a slightly slower full backup duration, and increased incremental data transferred per backup.
Do not enable Veeam’s Encryption feature. Although Veeam Backup & Replication supports encryption, for best performance, use the hardware encryption provided by the DXi appliance
When using the Veeam Linux Repository:
Quantum recommends increase the call execution timeout.
See https://www.veeam.com/kb1176 for more details.
When using Veeam Shared Folder repository:
Quantum recommends Disable SMB Server Signing (Server 2008/2012):
SMB Server Signing is disabled by default on the DXi, but must also be disabled on the Veeam server. (This can be changed on the local machine through a Registry setting, but if the server is part of an AD environment, the group policy must be changed to disable SMB Server Signing for the Veeam server for the change to be permanent.)
To disable Server Signing on the local server, disable (set to ‘0’) the following Registry values:
• HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\LanManServer\Parameters \RequireSecuritySignature
• HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\LanManWorkstation\Parameters\Requi reSecuritySignature
If the Veeam server is part of an AD environment, the group policy must be changed as well:
• Open the Group Policy Editor, and right-click-and-edit Default Domain Controller Policy.
• Go to Computer Configuration > Policies > Windows Settings > Security Settings > Local Policies > Security Options.
• Set "Domain member: Digitally encrypt or sign secure channel data (always)" and "Microsoft network server: Digitally sign communications (always)" to Disabled.
For more information, see https://technet.microsoft.com/en-us/library/cc731957.aspx
DXi backup I/O Guidance and performance expectations
Reverse Incremental Backup
During a Reverse Incremental Backup, Veeam injects changes into the .vbk file and then rebuilds it to the most recent state of the VM. This will cause simultaneous read and write operation on the DXi.
These simultaneous operations can negatively impact performance, causing an I/O performance bottleneck on the DXi. To avoid this situation, if you are using a CIFS NAS repository, consider running a standard incremental backup and then using the Active Full Backup method, to prevent I/O bottlenecks within the storage device. Multiple Full Backups will not consume as much space on DXi storage, because all similar blocks will be deduplicated by the DXi. If you are using the Veeam Linux Repository, consider running standard incremental backups followed by a periodic synthetic full backup.
Synthetic Full Backup.
This should only be attempted when the Linux Repository is employed on the DXi. The network overhead of transferring the data to the Veeam Backup and Replication server to synthesize is not necessary in this case, and the operation takes place entirely on the DXi. For this reason, this method is a good candidate for environments that have limited network bandwidth. It is not advisable to perform synthetic full backups when a CIFS Shared Folder repository is employed.
Instant VM Recovery
Booting virtual machines from the DXi repository is similar in performance to the Reverse Incremental and Synthetic Full Backup methods. However, this restore method could negatively impact performance, causing an I/O bottleneck at the DXi. This should only be attempted when you are employing a Linux Repository on the DXi.
The DXi works well with the Instant VM Recovery method, although it is not designed as primary storage. Thus, Instant VM Recovery should be viewed only as a temporary solution, until primary storage is available. It’s advisable to only start as many instant recovery sessions as necessary, because each session consumes memory on the DXi while active. Memory consumption varies, depending on what type of activity is present on the hosted virtual machine. It’s also advisable to redirect written blocks to a local high speed datastore to take advantage of caching. We also suggest that you migrate instant recovered virtual machines to permanent storage after instant recovery is complete.
Quantum best practices guide
Please be aware that starting from September 2018 downloading updates will require an active contract for the corresponding product.