Sizing for cloud VM backup operations
Backup and restore operations for cloud VMs depend on the same factors. Only the direction of data flow changes. Together, these factors determine the throughput of backup from snapshot and restore jobs:
The network bandwidth available to the NetBackup Snapshot Manager VM instance in AWS and Azure, or the I/O bandwidth available in Google Cloud Platform (GCP) and Oracle Cloud Infrastructure (OCI).
The compute resources, including CPU and memory, of the NetBackup Snapshot Manager instance.
The compute resources, including CPU and memory, of the storage server, such as MSDP. For MSDP-C, these factors also include cache storage capacity and IOPS.
The data path functions as a pipeline. The component with the lowest effective capacity limits overall throughput. Increasing capacity at one stage does not improve overall throughput if another stage remains the bottleneck.
The backup window, approximate change rate, and required operational headroom determine the number of parallel streams needed to maximize resource utilization and data-transfer throughput.
To sustain a backup from snapshot throughput of approximately 1.2 GB/s, some environments have provisioned the NetBackup Snapshot Manager and MSDP instances so that each supports approximately 1.2 GB/s of effective bandwidth, with approximately 16 vCPUs per instance. In these environments, scheduling approximately six or seven jobs in parallel has been sufficient to use the available bandwidth fully.
You can scale higher or lower throughput targets proportionally. Actual requirements vary based on workload characteristics, storage performance, and network conditions. Include adequate operational headroom when selecting the final configuration.