Performance optimization of backup jobs in Multi-Tenant environments

This article documents how to alter the performance characteristics of backup jobs created in the Veeam self-service backup portal (for VMware Cloud Director and vSphere) and jobs pointed to the Veeam Cloud Connect repository.

In a multi-tenant environment, one of the most important goals a Service Provider has is to ensure the availability of the resources to all users serviced by the environment while working to mitigate the “noisy neighbor” or “one-tenant-takes-it-over” issues. Based on this principle, the backup jobs created in Veeam Self-Service Portal and Veeam Cloud Connect are by default given an equal opportunity to process data within set allocation windows. This technique helps Service Providers reduce and ultimately eliminate the impact of backup data processing on the underlying infrastructure and other Tenants. In some scenarios, the Service Provider may choose to temporarily remove this restriction to improve the processing speed of the backups.

To improve processing speed for the backup jobs created in Multi-Tenant environments, the Service Provider can switch to the “Hybrid” (recommended) or ”Dynamic” modes of quota allocation.

After altering quota allocation mode in an environment, closely monitor backup infrastructure resources consumption.

 

This registry value created on each Veeam Backup & Replication server managed by the Enterprise Manager server (providing the Self-Service Portal) or Veeam Cloud Connect server itself.

Key Location: HKLM\SOFTWARE\Veeam\Veeam Backup and Replication
Value Name: CloudConnectQuotaAllocationMode
Value Type: DWORD (32-Bit) Value
Value Data (Default): 0
Acceptable Values: 0 | 1 | 2

 

0 – Default

Each IO stream for a backup file write operation is allowed to write at most 512MB of data every 15 seconds. This process gives each write operation equal access opportunity when using slower storage.

1 – Dynamic

The Target Data Mover responsible for the data processing requests the quota allocation based on the availability of the current resources combined with the resources needed to process the data sets more efficiently.

Note: High demand for the processing resources may result in a performance impact in slower infrastructure environments. Performance monitoring is advised.

2 - Hybrid (Recommended)

Defaults to the Dynamic quota allocation and resorts to the Default quota allocation as needed.