Is there anyone out there that can give me an idea about the throughput of VM restores on a vSphere còuster?
I backup VMs via vStorage API at a decent rate, between 30 and 45 MB/s via SAN.
I'm doing an alternate restore on a different vSphere cluster, authenticating on Virtual Center, and I haven't seen anything better than 3 to 4 MB/s via SAN. I'd expect to see something around 20 MB/s, am I missing something?
Tapes are not responsible, as I also tried to duplicate images on basicdisk, changing the Primary copy of course, and the restore thoughput is the same.
The bptm log is filled with "write blocks [...] writing <n> data blocks of 512" lines. Why 512?
I also modified the SIZE and NUMBER of *_DATA_BUFFERS, no change at all.
Have you tried to run the restore directly on the ESX host instead of VC?
If using ESX server, you can restore to a data store visible to that particular host. Hence your visibility to vSphere cloud is narrowed. However there is a distinct advantage in using ESX server in stage 2 especially if you are using SAN transport. ESX server is the real work horse allocating blocks for NetBackup to write; hence the restore performance will be good. Bypassing vCenter for restores is always a good idea.