Sizing Requirements for Disaster Recovery


Sizing requirements provide a best-guess estimate for the components required to maintain the Disaster Recovery solution.

Note that sizing requirements do not guarantee specific performance values. Note that required component sizes depend on several variables, and sizing requirements are merely estimates.

Component Sizing

Each VSA instance uses either its physical machine core or a vCPU to process each VM disk and 2 GB of memory to process the stream of data. Multiple disks and VMs are processed, depending on the availability of CPU/vCPU, the availability of RAM, and throttling settings. Performance is driven by the storage transfer medium, the disk speed, and the network bandwidth. You must account for these factors.

If you use the virtual appliance for deduplication, indexing, and other operations, refer to the MA hardware specifications in the Commvault documentation and upgrade the sizing, as needed.

Assume the following number of cores and the amount of RAM for each component in the specified time period:

  • Virtual Server Agent (VSA): 1 CPU core with 2 GB of RAM can process up to 4 TB data in an 8-hour window.

    This assumes that the necessary disk speed to read and write data and the necessary bandwidth are available.

  • MediaAgent (MA): 2 CPU cores with 16 GB of RAM can process up to 50 TB in an 8-hour window.

    This assumes that the necessary disk speed to read and write data, the SSD space for the deduplication database (DDB) and indexing, and the required bandwidth are are all available.

  • CommServe server: 8 CPU cores with 32 GB of RAM can run up to 1000 VMs.

Replication Sizing

Assume the following components and factors:

  • VM size (s) = 100 GB of RAM each

  • Data change rate in percentage (r) = 10%

    Note: This assumes a high rate of change. Typically the change rate is 2%–5%.

  • Bandwidth (b) = 10 GBps

  • Recovery Point Objective (RPO) = 60 minutes

To meet an SLA (recovery time) of 60 minutes, the replication operation must complete 10 TB (10% of 100 TB) in 60 minutes. With 10 small access nodes, 4 vCPUs and 8 GB of RAM are required to read data from the source site and transfer it to the secondary (DR) site in 30 minutes, if the necessary storage speed and bandwidth are available.

After the data transfer is complete, you must account for the time it takes to reconstruct or apply the changes on replicas in the secondary site.

With a 10% change rate, as per the example, with 10 access nodes needed, each access node should have a minimum of 2 cores and 4 GB of RAM to apply 10 TB of data changes to replicas in 30 minutes.