I had my first virtualization “a-ha” moment about five years ago when I first ran a copy of VMware Workstation on my laptop. The window flashed black, and as I watched the familiar BIOS boot sequence, I figured out just about all I have needed to know about how virtualization systems work: it makes a little computer in there, complete with memory, BIOS, and what looks like regular devices.

It’s that latter bit — the ability to share physical devices and substitute disk files for CD-ROMs and hard drives — that quickly led to my second “a-ha” moment: hey, that “hard drive” I added to that vm is just a big file on my computer.

As you know, everything that makes a computer unique is stored on disk: the operating system, applications, configuration settings, even the dreaded Windows registry — usually in hundreds or thousands of files scattered (and hidden) all over your disk. But when that “disk” is just one big file on a physical volume, suddenly it doesn’t seem so hard to back up, copy, clone — even version control — whole computers at once.

Virtualization, SAN, and software vendors offer a lot of bells and whistles — snapshots, real-time replication, volume mirroring, differential backups, etc. — but it is the “atomic” nature of virtual machines which make them so easy to back up and restore. And that sounds like the beginning of a disaster recovery strategy, doesn’t it?

In his recent Yankee Group Report Virtualization Eases Disaster Recovery, Zeus Kerravala argues that virtualization technologies bring full-fledged disaster recovery plans within the reach of SMBs.

Think about it — no business is too small to copy some (large) files offsite. As we march towards Anywhere, broadband connectivity will only become more widespread and affordable. ($1000 per month buys our office a 100 Mbps Internet line. Remember when a 1.5 Mbps T-1 cost more?) And when you can already rent storage by the month from Amazon S3 for $0.15 per GB, it’s hard to argue that a basic DR plan is out of anyone’s reach.

Consider a secondary site

Zeus goes a step further and recommends SMBs consider including a secondary site in any DR plan. Setting up a secondary site “in the old days” (i.e., pre-virtualization) usually required purchasing identically configured hardware to hold your data and to run your mission-critical apps. As a result, few projects to build secondary sites ever made it past the cost estimate — even among larger enterprises.

But the increased server utilization made possible by a serious virtualization program can free up decent hardware to be used in a secondary site. Your mileage may vary, but when Yankee Group undertook our crash virtualization program last year (“Project Shrinkray”, prompted by the move of our Boston headquarters), we were able to free up more than half of our production servers. And the footprint of our remaining servers was small enough that we saved money by moving them all to a nearby AT&T data center rather than building a server room in our new office. It also helped that our new home sits on a SONET ring — the Anywhere Network will come via whatever access technology makes sense.

It’s funny. It seems that whether we need a secondary site for a DR plan or a new order management system, we CIOs are always running into the same, familiar question: should I build it or buy it?

Option 1: Build It

Assuming you have (or can free up) enough servers to meet your computing requirements, I would argue that you’ve cleared the biggest cost hurdle. While data center space isn’t getting any cheaper — though this may change with falling energy prices and the sudden economic slowdown — other factors are working in your favor:

1. Free (as in beer) virtualization software. Microsoft’s entry into the virtuallization market is pushing prices down, as we warned earlier this year. In response to this increased pressure on the low end, VMware is now giving away its ESXi Hypervisor. You would still need to pay through the nose for extras like VMotion, but I’d skip them for your secondary site. Or you might opt out of the VI/3 family entirely for VMware’s long-free Server product running on, say, a CentOS Linux server. (And don’t forget open source virtualization alternatives like Xen which ships with Red Hat and SUSE Linux, if you’re so inclined.)
2. (Almost) free storage. Newegg, my personal choice for personal purchases (sorry, no corporate P.O.’s), is selling 1.5TB Seagate SATA drives for $130 (with free shipping . Or be conservative and pay $180 for the 1TB “enterprise” variety rated for 24×7 operation. How many do you really need to hold all of your live data?
3. “Offsite” may be closer than you think — like under your desk. If your critical systems are already sitting in a safe, professional hosting facility (and they should be), consider using your office as your “secondary” site for these systems. Or better yet — a branch office.

Option 2: Buy It — Anywhere

I enjoy fiddling with server guts at least as much as most IT guys, but sometimes I just don’t have the time. Cloud computing is the latest class of offerings being made possible by the emerging Anywhere Network. In Considering Cloud Computing for the Anywhere Enterprise, Carl Howe segments the cloud computing landscape and presents these offerings as a way to jumpstart your company’s evolution towards an employee-centric Anywhere Enterprise. Just google for VMware hosting providers or refer to VMware’s “vCloud Service Provider” list to find “Infrastructure as a Service” providers who will be happy to take all those little files — I mean servers — off your hands.

And don’t forget about DNS. If you do ever need to cut over to your secondary site — and assuming at least some of your critical systems are publicly available — a web browser is all you need to make the required changes with a hosted DNS provider like my favorite DynDNS.com. You may already have this capability (for free) where your domain is registered. Don’t forget to add this login information to your DR documentation.

Which option is right for you?

For my money, it’s hard to beat a couple of monster machines slapped together from spare servers and giant SATA drives. But you and your IT team must have the skill-set — and interest — to build it and maintain it. One quick test: did you click on any of the links to Newegg above? Have you heard of Newegg before? What about your team? If the answer to these questions is “no,” implementing this solution probably won’t be as much fun as it sounds.

From an accounting point of view, computer upgrades should be capitalized and depreciated accordingly. If your financial circumstances favor capital expenditures over opex spending, that could be reason to lean towards building your solution — and don’t forget that you should also capitalize an outside consultant’s fees to help with the initial implementation. If you choose a services-based approach, now is the time to account for these expenses in your 2009 budget.

However you account for it, virtualization — along with cheap storage locally or in a cloud on the Anywhere Network — has eliminated much of the cost and complexity holding most of us back from implementing a workable disaster recovery plan. So now what are you waiting for?