Find out more about our consulting services. McLean. Boston. New York.
There are few things more alluring to a system administrator than the promise of easy configuration and worry-free maintenance. It is the promise of many cloud-based service providers – providers with big names, like Amazon with its EC2 service, and Microsoft with Azure. The positives may seem downright impossible to ignore when it comes time to write the next hardware refresh budget: managed upkeep; paying for only the hardware and bandwidth you use; the ability to scale your infrastructure with a few moments notice. Best of all, those frighteningly complex architectural diagrams with redundant failovers, duplicate NICs piped through separate channels and backup domain controllers all get replaced by a big fluffy icon with a label of “Not My Problem Anymore”.
But when parts of Amazon EC2 went down last month, taking popular sites like Reddit and Foursquare down with it, the original meaning of that cottony cloud symbol hit home for those looking to dip their feet in the cloud computing pool. The cloud represents the nebulous and uncertain nature of the processes that take place in that realm, processes that can be influenced but never fully controlled. Even when the vendor promises fail-safes against widespread outages or rattles off a SLA that has a service uptime starting with 99 and keeps going past the decimal point, that’s no guarantee against reality. Keep in mind the old stock trading caveat: “Past performance is not an indicator of future results,” a sentiment that applies equally well to service uptimes.
Don’t let the big names behind those cheery cumulonimbi blind you. The 2000s are gone and we should have learned our lesson by now: too big to fail does not apply to anything in life – tech especially. How often does your Gmail go down? Not very often, right? But you remember that last time it did, and maybe you finally clicked the “Offline” mode setting that saves a local copy of your mailboxes so the next time it did you could still get all that information you took for granted as living permanently in the cloud, accessible from wherever you are at any time.
The problem with that little picture of a cloud comes back to the concept of a single point of failure, the bane of every network architecture — no matter how well designed. At NavArts, we’ve seen our share of over-engineered solutions loaded down with compound buzz words — load-balanced, round-robin, off-site, cold-backup — that all still depend on a single SQL database to provide their functionality. And it’s not just the technology that serves as a fail point. There’s the human angle: a single administrator, working in an immaculately designed system that is so complex that the user bypasses the entire process just to push out a small change to their site for a typographic emergency.
So should you avoid cloud computing? Absolutely not. Cloud computing can certainly be the right fit for your organization, providing all the promised benefits at a reasonable cost. But you should plan for its usage and do your homework before you take the plunge. And most definitely avoid the hubristic impulse to label that figure on your diagram “Not My Problem Anymore” — just in case of a rainy day.