Difference between Backp & Replication (and how to restore)

[pkelly_sts]

I've been using Veeam for backps for around 9 months now and have it deployed in 3 sites. On 2 of those sites I have it working nicely with Direct SAN access so backups run at plenty acceptable speeds and all is well.

I'm now looking to architect a solution for a (4th) DR site with the potential to, at least backup but possibly also replicate 2 sites to this DR site, whilst also cutting out the need for removal of tapes off site on a daily basis.

My grand plan, as we are quite spoilt with available hardware is along the following lines:

#Main site#
Primary site with the most VMs (around 35) running on 5 hosts with around 8Tb of FC-based VMFS volumes provisioned to the hosts.
10 x Enterprise B&R CPU licenses
Single B&R server with R/O FC SAN access backing up to both internal storage (15Tb) & external (and different) FC storage (20Tb)
MSL2024 with 2 x LTO-4 tape library SAS-attached to B&R Server dumping Veeam backups to tape nightly
MSL4048 with 2 x LTO-3 SCSI tape drives currently not doing a great deal
100Mb P2P link to DR site
Daily incremental change rate of around 100Gb, synthetic fulls on a Friday night/sat morning
14 restore points retained
Tapes taken offsite daily by Iron Mountain & held for 1 week before being returned.

#DR Site#
An ever-decreasing footprint of active users / VMs so current aging infrastructure being replaced with DR kit
Current remaining VMs being backed up (badly!) by vRanger
10/15 miles away from main site
2 "Spare" servers with FC connectivity ready to deploy as DR Host servers
6 x Enterprise B&R CPU licenses
60Tb (useable) of FC & iSCSI RAID-5 SATA storage for DR/Backup purposes ready to deploy
2 x FC SAN Switches
MSL6030 tape library with 2 x FC-attached LTO-3 FC tape drives
100Mb P2P link back to main site
Tapes being taken offsite daily by Iron Mountain, unsure of retention period for now

#3rd Site#
Smaller version but similar setup to main site, i.e. 3 hosts, FC storage, 6 x Enterpris B&R licenses
10Mb MPLS connection to DR site (so not as large a pipe)
Around 15Gb of daily change rate incrementals

My "grand plan" (discounting Veeam 7 features for now as not quite yet available) is roughly as follows:

Continue backups on main site, write them to tape but retain tapes in library and simply let them rotate through automatically. No further off-site removal of tapes
Run a second backup directed to the DR site, Synthetic fulls will run at DR site as well, thus managing retention automatically
Tape backups at DR site, tapes left in library, rotating automatically as at main site (LTO-3 would be slower but not really an issue as plenty of time to complete with no impact to live service

This leaves me with the following:
2 copies of backups on-disk in 2 locations
2 copes of backups on tape in 2 locations
Optional "Monthlies" to be retained on each site, removed from library but kept locally, so 2 copies of monthlies retained for however long required.

So, backup-wise I think I'm well covered, but I'd like to be greedy & make use of the hardware (and licenses) we already have to have a full replication DR solution as well.

What I don't yet understand is how/where replication jobs write to. I know that the B&R server has R/O access to VMFS and have tested a "full VM recovery" quite nicely, knowing I can svMotion a restored VM back onto primary storage. However, in the event of a DR I don't think it's realistic to expect to spin-up more than a few VMs at a time out of backup storage and start svMotioning it at the DR site.

Is replication somehow written to DR site VMFS (via proxy appliance) or is it written to a B&R repository?
If written to a repository, how do replicated VMs "pop into life" at the DR site currently if we invoke DR?
If we're replicating to the DR site anyway, is it more efficient to just locally back-up the replicated copies of the VMs?

I know V7 will significantly change some of this anyway (no need to run duplicate backup jobs, just copy backups to an off-site repository with retention plan) but need to do something sooner that I'll be able to deploy v7 which is sounding like a Q3 release (which could easily be a late Q3 release).

Any suggestions, or pointers to similar discussions/solutions would be greatly appreciated!

Regards,

Paul

[foggy]

Is replication somehow written to DR site VMFS (via proxy appliance) or is it written to a B&R repository?

Paul, replication job creates an exact "ready to go" copy of a VM on the specified target VMFS datastore, registers it on the target host and maintains it in sync with the original VM so you could be able to failover to it quickly (by simply turning it on) in case of DR.

If we're replicating to the DR site anyway, is it more efficient to just locally back-up the replicated copies of the VMs?

This is a viable approach, however it has several peculiarities that worth to be mentioned. More details can be found in this topic.

I suggest reviewing the corresponding user guide section (p.50) as well as this webinar (Session #3) to learn about replication architecture and functionality.

Hope this helps. If you need any further clarification, feel free to ask here.