The Cranky Admin
How Much Cloud Is Too Much?
Don't let fear force you to spend more than needed on backup and disaster recovery.
I recently sat down with a customer to redesign their disaster strategy. After being properly terrified by a cold call from a local VAR, they were gung-ho about the hybrid cloud. They needed proper disaster recovery as a service, you see, and could I provide that for them cheaper than the VAR in question?
This seemed to come out of left field, so I was left to wonder what made this normally penny-pinching customer so ready to part with their money. With some work, I realized that the general sense that the world is spinning out of control and nothing is sacred anymore had gripped my customer.
And here I thought that was only affecting me.
I got the numbers the VAR had quoted to the customer, and we ran a thought exercise with them. Right now, today, the on-premises IT apparatus that the client has is pretty robust. A lot of bits have to fail for the servers to go kablooie. Everything is backed up regularly off-site to a backup box in a colo about a half-hour drive away. If worst comes to worst, we can physically go down to the colo, retrieve the drives from the backup server and build the customer a new IT apparatus within a week.
Barring some improbably freak accident, anything that could kill the customer's on-premises datacenter would also destroy everything else in the building where that datacenter is located. If that happens, the rest of the business is hooped anyways.
This customer has multiple sites all across the country, but all they do is funnel data and samples back to the one central location where a bunch of nerds bang on it. Those nerds need some reasonably specialized gear to do their work, and replacing a bunch of customer workstations, mass spectrometers and so forth will take rather a lot longer than it takes me to rebuild a 100-host datacenter from backups.
Virtualization is a wonderful thing.
The customer pointed out that if he had a proper disaster recovery solution and something truly bad happened -- say, for example a hail of missiles obliterated his central location (not to mention him, his HQ staff and me) because it's parked next to the local military base -- that folks in one of the other cities would be able to continue working. This is a great point, but the real question is "what good would that DR-enabled IT do them?"
Mind-Blowingly Simple Restoration
At this point, I blew the customer's mind by showing him just exactly how simple it was to restore his infrastructure from backups. Making the computers go really isn't the hard part of this.
If he wants to spend money and truly prepare for disaster, he should be sinking money into knowledge transfer: i.e., making sure people in other locations know the business processes required to do something with the data and applications once they're available. Maybe move some of the pricy non-IT equipment to other cities and generally make sure the other pieces of the business are resilient.
Getting data off premises in the first place is usually miserable. Most customer-facing Internet connections are punitively expensive. But once in a colo or in the cloud, it's cheap to keep two copies in two places, and most customers just don't need the ability to flick a switch and make it all go, nor to have it multi-site-cloudy-super-resilient with extra expensive on top.
You'd be surprised how little it costs to just put a local IT consultancy in your backup city on retainer and pay them to restore from backups in the unlikely event everything goes to hell. For many organizations, that's more than good enough, especially in the chaos that would ensue from having the majority of experienced backend staff taken out of the equation one way or another.
I don't want to knock the cloud, renting gear in multiple colos, or any other multi-datacenter strategy you want to entertain. If you really feel that's a good plan, then by all means go hard.
I'm not here to blaspheme against the almighty cloudy religion, but I fear this is going to be the new trend: disaster porn turned into sales pablum and forced down the throats of the technically uninitiated. Selling backups, DR and cloudy resilience is easy if you can tap into primal existential terror.
Moving everything to the cloud might provide a seemingly quick way to get that multi-datacenter feeling, but it's important not to be driven by an unsettled feeling of panic and loss of control brought about by political anxieties. Don't let existential and abject fear -- something that, sadly, many salespeople seem to be gleefully trying to capitalize on -- override real technical concerns.
Backups are essential, and DR is nice to have, but don't let the suits be bamboozled into buying more than they actually need.
Trevor Pott is a full-time nerd from Edmonton, Alberta, Canada. He splits his time between systems administration, technology writing, and consulting. As a consultant he helps Silicon Valley startups better understand systems administrators and how to sell to them.