As we discussed here earlier, there are 3 transport modes to get data from vSphere for backup.
These modes are also available for restore. By default, first mode (order: SAN, Hotadd, NBD) that meet requirements, is selected for restore. For SAN direct mode, a requirement is thick provisioned disk type. You can select each available type (thin, thick eager and lazy zeroed) in restore wizard.
My recommendation: If you want to leverage SAN direct mode, choose Thick eager zeroed! This option is much faster than lazy zeroed!
I tested these settings in different environments. For example, I saw differences of
67 % (150MB/sec for eager vs. 100MB/sec for lazy),
Please share your experience with SAN mode restore! I am interested in observed bandwidth comparing backup with restore.
Thanks @vNote42 , i already read a san restore in the website best pratices (i don’t re found the article ) It’s a good reminder!
I dont have any numbers but especially on old HP EVA storages eager zeroed restores were much faster then any other SAN mode. Which devices did you use in your tests @vNote42 ?
I dont have any numbers but especially on old HP EVA storages eager zeroed restores were much faster then any other SAN mode. Which devices did you use in your tests @vNote42 ?
Latest arrays I tested with: HPE 3PAR (FC), Primera (FC), Nimble (iSCSI and FC). I also think the version of vSphere matters here. When you say EVA, the version number was probably <5
Didn't EVA go up to 5.5 or even 6.0? Can't find them in the VMware HCL anymore
Which SAN was used in your second example? The difference is like day and night...
Didn't EVA go up to 5.5 or even 6.0? Can't find them in the VMware HCL anymore
Which SAN was used in your second example? The difference is like day and night...
93 % difference I saw on a Nimble HF40; FC host ports and Sync-mirror. But interesting as well: on this array I had nearly the same speed for backup and restore (eager thick) in SAN mode.
) It’s a good reminder!
