As I write this on a cold winter Friday afternoon, most of the Northeastern United States is abuzz in preparations for an incoming blizzard. By the time this posts, we'll all know whether the "snowmageddon" Storm Nemo hype spun by the weather folks was true. In the meantime, the forecast of multiple feet of snow combined with high winds has created the real possibility of long-lasting power outages and tough travel. As a result, IT directors throughout the region are scrambling to make sure they're ready.
Of course, the lead-up to a big storm is always the time people seem to realize they've forgotten a detail or two. As a consultant and adviser, I've received more than a few panicked calls over the years asking whether I happen to have a few extra servers (or even an enterprise-class SAN!) kicking around "just in case." We all know that planning for a disaster when you're staring it in the face is probably too late, but what if there's a different way to approach the question of disaster preparedness in IT?
In the old days, the idea of transitioning mission-critical workloads to a data center that's not in the path of a storm might have been a capability that only a few of the very largest enterprises could afford to implement (and even then for relatively small portions of their infrastructure). However, as storage and virtualization technologies have continued to advance and cloud infrastructure offerings mature, the ability to completely sidestep disasters rather than weathering the storm has become available to many more organizations.
Not all that long ago, if you wanted to make sure your IT operations could avoid a major weather event, you'd need a hot site. Typically, that meant full-scale data center facilities, a copy of nearly every piece of equipment you run at the primary data center, complicated failover procedures, and substantial network facilities to support cross-site data replication and post-failover user access.
Anyone who's done that work will tell you that, unless you're huge, building your own enterprise-class data center is often about as wise as investing in a boat -- except without the occasional joy of fishing trips when it's not in for repair. Raised floors, racks, and electrical distribution are one thing, but dual-redundant generators, UPS, and HVAC can add up to a prohibitive price tag.
Don't build your own redundant data center -- rent one instead
Fortunately, the dot-com boom in the late 1990s saw the construction of loads of massive data centers and the rise of the colocation industry. No need to build your own data center -- rent a tiny piece of someone else's (increasingly cheaply, I might add) instead.
On the equipment front, the tens and hundreds of servers with direct-attached storage of yesterday have largely been replaced by virtualization hosts attached to shared storage. What used to require a myriad of host-based software to replicate and keep in sync now might only require SAN-to-SAN replication and a piece of software to automate the failover of the virtualization environment. To be sure, none of that is cheap (especially to small businesses), but it's certainly not the multi-million-dollar adventure it used to be.