We're looking to get a new SAN, and so we're re-examining our storage, network and backup issues. Our plan is to get an EMC VNXe 3100 with SAS drives for production data, then buy an expansion shelf and populate half of that with SATA (6 TB raw) drives. Our backup software will then use the SATA as the target.
Our initial sales rep said that would work fine, followed by another EMC sales rep (not coincidentally trying to sell a backup storage appliance, Data Domain) saying that you should never ever put the backup set on the same storage. I'm sure that would be better, but budget is a concern for us, and using the SATA seems a reasonable compromise, but I'd like to be sure.
In my mind, the disk backup target on SATA is to give us 1.) smaller backup window, and 2.) convenient tapeless recovery. We'll still be using tape for disaster recovery and using the backup software to move the disk backup to tape. We only have one site, so replication isn't a great option. Since budget is a concern for us, this seems reasonable as opposed to buying a special device.
The only failure mode I can think is if both controllers on the VNXe fail simultaneously, we're not going to be able to get to either production disk or disk backup. But the odds seem really low for that. Seems that if that happened, we'd be looking for the tapes anyway, because the server room would be gone.
So, am I crazy? Is there anyone else doing something similar?
While your proposed solution would work, as Chopper3 mentioned it's not ideal. We use, and have had great luck with Nexsan SATAboys for disk-to-disk backup. The SATAboy isn't going to give you all of the advanced features that your EMC will, but honestly you don't need that for backup staging. For that purpose, you just need a bucket-o-disk, which the SATAboy will give you. They're very reasonably priced as well - a lot cheaper than a purpose built backup appliance. If you went with the SATAboy, you'd just attach it to your existing backup server via Fibre channel or iSCSI.
One of the reasons you mention for being hesitant to let your production data and backups co-habitate the same SAN is: how do we get our backups back if the whole SAN goes down. That is indeed a good reason to separate things out, but there's another - IOps. With both workloads on the same SAN (even if not on the same disk array), you'll be pumping twice the IOps through the SAN as you would if the backup staging was elsewhere. That may or may not be a problem depending on how loaded the SAN will be.
So, just wanted to throw that option out there in case you're able to scare up a bit of additional money for your backup environment.
Disk-to-disk backup can give you some benefits:
What it won't do, however, is protect you from
In my experience, user error is by far the number one reason for a disaster recovery, followed closely by HDD failure. So disk-to-disk backup will protect you against that. The number 3 reason, though, is server failure -- e.g. power supply, motherboard, OS -- and disk-to-disk will leave you very vulnerable in that scenario.
If these risks are acceptable to you, then disk-to-disk may be a viable solution for you. However, if your data is important enough to require surviving any of these other disasters, off-site backup is mandatory.
Also, a two-tiered backup system may be an option. For this approach, make a disk-to-disk backup, and then make an additional (tape, etc.) backup that goes off-site. This gives you the ability to quickly and easily recover data from a HDD failure or user error/PEBKAC, while also giving you the security of off-site backups allowing you to recover from just about anything short of nuclear war. (This is what I use for backups, although for years I relied solely on disk-to-disk and accepted that recovery could be as involved as plugging in my disks into an entirely separate system and recovering the data from there.)
You do say you're using tape backup in addition, so it sounds like you are going for this two-tiered approach. In which case I say no, you're not crazy. However, it appears from your question that you may not have considered the risk of an OS or other server failure bringing down the entire box, although if that's your only production server then having a physically separate backup device won't really dig you out of that hole anyway.
Certainly if you have the budget then you should go for your short term backups on a separate array but better to have some backup, even if one the same array, than none.