But I do, of course, talk to lots of IT folks for the articles I write (and many of my friends are in IT). And, like anyone who owns and uses computers as a business tool, I periodically make quick rough "is it worth it" calculations on whether a given purchase makes sense, in that often-fuzzy balance between frugality and productivity.
For example, replacing my desktop computer a little over two year ago was, once I acknowledged the realities, a no-brainer; my then-main system was running slow, adding memory wasn't fixing it, and while an OS reload might have done the trick, the systems was five years old, so putting any more than about $50 to $100 in hardware was a bad investment. (See my blog post, Fix One, Fry One, Buy One: Death Of A Motherboard.) Amortized over three years, a new computer cost about a buck-fifty to two dollars per working day... and just the wait time due to system sluggishness was way more counter-productive than this amount. Way more.
Similarly, two reasons I moved from recently from a "dumbphone" to a tetherable smartphone were checking email while away from my office, and Internet access for whatever notebook or netbook I might be toting -- it doesn't take much per month to recoup the price of 3G data service and the tethering option.
Ditto having a UPS, which, at $100 to $200 for desktop protection, better be a no-brainer.
For the server-oriented Windows applications that businesses need, like Microsoft Exchange, Active Directory, SharePoint, fileserving, BlackBerry Enterprise Server (BES), downtime has become anathema, even in small-to-midsize companies.
Some preventative measures are affordable; for example, a server-class UPS, RAID or other redundant storage (see Data Robotics Announces DroboPro FS Storage Appliance).
What about the server hardware? Having a spare machine doesn't have to be that expensive. But using it to provide high availability (HA) or non-stop (fault tolerance) can be -- or at least, has been -- pricey.
Prices for a fully fault-tolerant system from Stratus, for example, begin around 14 grand for an entry-level quad-core single-socket machine... and that doesn't include the operating systems. If "high availability" will do, Stratus offers its Avance high-availability software, which costs around $5K for license to use with a pair of x86 machines. According to Stratus, "Avance software runs in the "four-nine's plus range," and is designed to prevent failure and data loss -- about 80% of what you get from fault tolerance.
Or you could go cloud. Geminare, for example, offers cloud-based server failover (see my InformationWeek blog post, Recovering Servers In The Cloud Affordable For SMBs). Geminare's Cloud Recovery service will run around $400 per server per month -- and that doesn't necessarily include the data side.
Marathon Technologies' new everRun MX software lets you create creating fault-tolerant Windows application environments using commodity Windows boxes (see my October 5, 2010 InformationWeek news story, Marathon Unveils Low Cost Server Fault Tolerance, which offers another opportunity to do IT arithmetic.
everRun MX keeps servers in sync, so if one hiccups or dies, your application(s) keep on ticking, uninterrupted. For a pair of Windows servers, everRun MX will cost about $10,000. That's a one-time license cost, including the first year's support, and doesn't include the hardware, OSs, applications, of course.
For a small to midsize business, 10 grand can look like a big number -- big enough to be rejected out of hand as "not in the budget."
That's a one time price -- and it's for "two servers" not "two specific servers." (I checked.) So you get to keep using it as long as you've got a pair of Windows servers, even if you're refreshing/replacing the hardware every year. Assume five years, that's $2,000 per year -- less than $200/month. Say a second server is about $100/month (over three years), call it about $300/month for server apps that won't get whacked by hardware hiccups -- or even by a fifteen-minute routine, scheduled hardware downtime for maintenance.
And if you're using everRun MX with just a pair of servers, you can do load-balancing. (If you're using everRun MX to do "n+1" protection for, say, three or four servers, Marathon says that the plus-one machine will only be providing failover.)
$300/month to ensure server uptime is less than the cost of an hour or so of IT contractor time per week. It's likely way less than the out-of-pocket cost to restore service from any outage. Or the cost of not doing business for an hour.
Like I said, I'm not an IT person, or an industry analyst. So I'm sure there's more complicated figuring that needs to be done. But sometimes simple is a good place to begin.
How much is keeping server applications up worth to your business?