Planning backup storage

Backup storage requirements depend on the Backup Gateway destination. Backup data can be stored in a local Virtuozzo Infrastructure storage cluster, on an HPE StoreOnce device, on an NFS share, or in a public cloud, including S3-compatible storage, Microsoft Azure, OpenStack Swift, and Google Cloud Platform. With a public cloud destination, Backup Gateway can also run in a virtual machine.

For every destination, add the Backup Gateway reservation from CPU and RAM requirements to the system and storage reservations of the node that hosts it. If the local storage cluster uses EC, also add the storage-client reservation on every node. Backup Gateway consumes the same CPU and RAM when working with local storage, HPE StoreOnce, NFS, or public clouds.

Choose a destination

Choose the destination before sizing the node. Local storage keeps backup data on premises and requires storage capacity in the cluster. An HPE StoreOnce device or an NFS share stores backup data outside the local cluster but still requires a local storage disk for internal needs. A public cloud destination uses local storage as a staging area before uploading data to the cloud.

Local storage cluster

Use a local storage cluster when backup data must remain on premises. A single-node deployment can be used only when you accept that a node failure may make data unavailable or cause data loss; in production, use multiple nodes and a host-level failure domain.

• For a single physical node with one system+metadata disk and five storage disks, using EC 3+2 with the disk failure domain, the service reservations are about 11.1 GB RAM and 2.7 cores, including about 2.6 GiB RAM for the EC storage client on a 12 GB node. The minimum hardware configuration is 12 GB RAM and 6 cores; the recommended configuration is 16 GB RAM and 6 cores.
• For five nodes with one system disk, one metadata disk, and ten storage disks per node, the base service reservations are 11 GB RAM and 3.7 cores on management nodes, and 8 GB RAM and 3.7 cores on secondary nodes. With replication, the minimum configuration is 12 GB RAM and 4 cores for management nodes and 8 GB RAM and 4 cores for secondary nodes. If the storage policy uses EC, add the EC storage-client reservation on each node and increase the minimum configuration to 16 GB RAM and 4 cores for management nodes and 12 GB RAM and 4 cores for secondary nodes. The recommended configuration is 24 GB RAM and 8 cores for management nodes and 16 GB RAM and 6 cores for secondary nodes.

HPE StoreOnce

Use an HPE StoreOnce destination when backup data must be kept on a StoreOnce device that runs the HPE Catalyst storage protocol. Backup data is stored on HPE StoreOnce, but the Backup Gateway node still requires one storage disk of at least 100 GB for internal needs and one metadata service on the system disk.

Each Backup Gateway node must have network access to the HPE StoreOnce device, and the device must have one or more Catalyst stores prepared.

For a one-node HPE StoreOnce destination with one system+metadata disk and one storage disk, the service reservations are 6.5 GB RAM and 1.9 cores. The minimum configuration is 8 GB RAM and 4 cores; the recommended configuration is 8 GB RAM and 6 cores.

NFS share

When the destination is an NFS share, backup data is not stored locally. Still, the Backup Gateway node must have one storage disk of at least 100 GB for internal needs and one metadata service on the system disk.

For a one-node NFS destination with one system+metadata disk and one storage disk, the service reservations are 6.5 GB RAM and 1.9 cores. The minimum configuration is 8 GB RAM and 4 cores; the recommended configuration is 8 GB RAM and 6 cores.

Public cloud

When the destination is a public cloud, the local storage disk is used for data staging. In a one-VM deployment with one system+metadata disk and one storage disk, using the 1+0 encoding mode with the disk failure domain, redundancy is provided by the virtualization solution.

• Backup Gateway uses local storage as a staging area and to keep service information before data is uploaded to the public cloud. For this reason, local storage must be persistent and redundant. Using temporary disks may result in data loss.
• Provide enough logical space for staging. For example, if backups run daily, provide enough space for at least 1.5 days of backups. If the daily backup total is 2 TB, provide at least 3 TB of logical space. The required raw storage depends on the redundancy mode.
• A separate object container is required for each backup storage cluster.

For this public cloud example, the service reservations are 6.5 GB RAM and 1.9 cores. The minimum configuration is 8 GB RAM and 4 cores; the recommended configuration is 16 GB RAM and 6 cores.

Virtual machine deployment

VMware vSphere is the only officially supported hypervisor for running Virtuozzo Infrastructure as backup storage in a virtual machine. Do not use other virtualization platforms, as doing so may result in data corruption.

• Use VMware vSphere 6.7 or newer and VM version 14 or newer.
• The host must have enough memory. At least 8 GB RAM is required for a node with one storage disk running Backup Gateway.
• The vSphere datastore must have enough free space for the system disk, storage disks, and the Virtuozzo Infrastructure template. The template takes about 35 GB, and the maximum recommended size for one virtual disk is 16 TB.
• Plan the size of virtual disks beforehand and reserve enough space for expected data growth. Virtual disks cannot be resized later, though new disks can be added.

Network requirements

To use only the Backup Gateway and storage services, configure two networks, for internal and external traffic: