As a general rule we recommend the following guidelines for deciding when to use WAN accelerator and for setting expectations for performance:
Global Cache on Spinning Disk
Link <3Mb/s - WAN likely saturated; processing rate dependent on data reduction ratio (estimated 10x)
Link >3Mb/s and <50Mb/s - WAN will not be fully utilized; expect ~5MB/s processing rate but less bandwidth
Link >50Mb/s - WAN will not be fully utilized, using direct mode copy will use more bandwidth but likely be faster
These numbers are to be considered as a base line , “Your mileage may vary”. The performance of the underlying storage where the Global Dedupe Cache is located can greatly impact the performance of the WAN Accelerator function.
Tests show that there are no significant performance differences in using spinning disk drives as storage for the target WAN accelerator cache rather than flash storage. However, when multiple source WAN accelerators are connected to a single target WAN accelerator (many-to-one deployment), it is recommended to use SSD or equivalent storage for the target cache, as the I/O is now the sum of all the difference sources.
WAN Accelerator Cache/Digest Provisioning
If we assume that we have 3 VMs, each with unique OSes (for instance, Win 2008R2, Win 2012R2, Solaris 10) each OS requires 10GB to be
allocated for it. The Cache itself is wholly independent
from the digests required. That is, the Veeam GUI does not make any determination of
how much you can allocate for a digest
and so on. The digest is essentially an indexing of what cached blocks go where. For digest size, 1TB of
VM disk capacity we are backing up should correspond with 20GB of disk space. That is, for 10VMs we are backing up whose capacity is 2TB, you
must account/allocate 40GB for digest data on the Source WAN Accelerator. This limitation is also applied to the Target WAN Accelerator.
For a Many-to-1 setup, the global cache is calculated per 1 Source WAN Accelerator working with the Target WAN Accelerator. Therefore, the
global cache needs to be increased proportionally. If we use the same VMs in the previous example, the cache is only required to be 30GB.
However, since we’re using 3 Source WAN Accelerators, the cache size must be 90GB in response. On the Target WAN Accelerator, not only is the
cache size dictated by the amount of Source WAN Accelerators, but so is the Digest on the target end—in this example, we require 120GB of Digest
space, which added to the cache size (90GB) results in requiring a 210GB volume size at a minimum