Time for the reign of clouds?

Ex-Google bod Dave Girouard writes of what business owners should consider when thinking of moving their operations into the cloud:

...one of the chief conundrums of cloud computing: you are powerless to fix a problem, and entirely dependent on somebody you can’t see, hear or yell at, to fix it. People hate that.

Girouard's perspective is interesting as he's a gamekeeper turned poacher; used to lead the Google Apps team whose entire raison d'être is cloud service hosting; he's now leading a startup and a customer of cloud services. So why does he think that a firm should host itself in the cloud?

He quotes and refutes three common arguments against cloud hosting:

1. "These big outages mean we should keep things in house"
   Most company in-house IT firms have no idea of the availability they actually provide. When you are trying to measure anything better than about three 9s (99.9% availability - fewer than 10 minutes downtime per week, or 1 in 1000 queries failing) you need to have a really good way of measuring availability from a user's point of view, which generally means several user-emulating bots probing the system and reporting on errors. The scale at which cloud hosting operates means that they have a much better idea of how close to, say, four 9s they are running at. Of course, they won't be able to measure directly any interruptions in the network between the cloud and an individual client, but they will have profiling of normal client activity and be able to spot unexpected traffic drop-off.
2. "I need somebody to talk to when a service interruption occurs"
   This is what really bugs me. When things go wrong in a company, and they will, the worst possible thing the affected workers can do is to badger the IT department with "when will it be fixed?" questions. As soon as the relevant infrastructure team recognises and acknowledges the problem, the best thing that the affected workers can do is to plan how to work around the outage, and eread the communications from the infra team that provide updated information on the scope and expected length of the outage. Repeatedly emailing or phoning to them is not going to help, no matter how diligent you wish to appear to your boss.
   One approach I've seen work well is for the infra team to designate one person to manage comms - respond to all incoming email, forward any critical information to the infra team actually solving the problem, and compose and send updates to the affected teams at designated times. The infra team may also designate a second member as a fixer-at-will to be given to affected teams for help with workarounds while the actual problem is not yet fixed.
3. "Cloud is OK for non-critical applications with non-sensitive data"
   Encryption is cheap. Encryption is (relatively) easy. Use HTTPS and only your employees and the hosting company sysadmins will be able to see your data. Encrypt critical information before sending, and the problem is solved - and as a side benefit, many fewer of your employees will have access to the secret data, if you do it correctly.

The most obvious argument against cloud is "what happens if the network between us and our cloud service provider is down / congested?" Clearly you have to be careful about your choice of ISP, and identify multiple redundant routes to your cloud service provider. This is an extra expense and hassle, but it may well still be worth it. Of course, it significantly increases the importance of building an IT department that can plan, build, measure and maintain network connectivity; this is harder to do, and harder to recruit for, than just paying Microsoft / Cisco / IBM for bundles of "enterprise" software and installing it on expensive hardware.

Girouard won't make any friends in IT with this though:

Further, this confused IT leader thinks his team can manage a service more reliably than a company whose entire existence depends on its ability to do so. To put it bluntly, Google has assembled the greatest collection of computer science talent in the world. Similarly Amazon has a multi-year lead in delivering compute power by the drop [...] Your IT organization simply doesn't rate at this level.

But it's (somewhat) true. Really good IT people don't generally stick around company IT departments very long; there are other, more tempting and lucrative gigs e.g. contracting. And if you rely on hire-at-will fire-at-will contractors for most of your IT department expertise, what exactly are your objections to cloud hosting? The number of times I've seen a single-point-of-failure IT admin depart, or threaten to depart, and leave his manager and teammates scrambling...

It will be interesting to see how the expertise profile and size of IT departments in companies migrating to cloud services will change. I expect quite a few false starts, but eventually an IT department's job (and an important one) will be focused around network connectivity rather than providing email, storage and distributed computing services.