If it only takes twelve steps to deal with an addiction, what are potential cloud customers to make of the first link returned by a query of “cloud migration” – 17 Steps to Cloud Migration?
With the notable exception of Larry Ellison, we seem to have a general consensus within the industry – from customers and providers alike – that the cloud can offer significant benefits over more traditional deployment scenarios. There are exceptions, of course, and the industry is not likely to replace its existing infrastructure investments wholesale with cloud alternatives. But the cloud is useful, flexible and powerful, and customers are interested.
Which is why it’s interesting that the industry spends next to no time talking about the most obvious barrier to entry: migration.
Most of the workloads we see deployed to cloud environments are net new. For every WordPress.com that bakes cloud into the infrastructure at a fundamental level there are ten shops use it, effectively, as a simple utility. Which is fine, and an excellent illustration of the benefits enjoyed by users of elastic, pay-as-you-go infrastructure. But the cloud could be doing more of the former, and one of the reasons it’s not is that it’s too difficult to get from Point A (local) to Point B (cloud).
It’s not that the technology is yet to be invented; live migration of virtual images is, at this point, a solved problem, and the cloud is on some level – PaaS fabrics aside – little more than a giant virtualization platform. And yet millions of potential customers continue to run cloud-appropriate workloads on local hardware simply because it’s too difficult to move. That’s a shame.
We see a similar inertial resistance to move amongst Microsoft customers, with both the Windows and Office products. In spite of considerable efforts behind various alternatives – Mac, Linux, thin client, SaaS and so on – the market has largely been unmoved. Some of that, of course, is due to Microsoft’s efforts to evolve the products, and some, the DOJ might argue, tactics less savory. But the point is that even if those platforms are exactly comparable, or superior, on a feature function basis, the friction from moving from the PC platform to something that’s not the PC platform is often enough to prevent the move.
The advantage with cloud workloads is that at least at in the Infrastructure-as-a-Service space, the required application changes are few. Currently running a standard LAMP stack? Not a problem: that can be replicated pretty seamlessly in the cloud, on whatever flavor of Linux you happen to run. Ditto for Windows. And for all of the ways the environment is unique, like the way you access the storage layer in the cloud, it’s far more similar than different. At least compared to the differences between running Windows and a Mac.
Still, the attention paid to packaging existing infrastructure and readying it for cloud deployment has been minimal. The idea of snapshotting or imaging a given server running production workloads, and migrating that – live or otherwise – direct to a public cloud infrastructure is still mostly a pipedream. There are a few layers to production server migration, of course: application and data, for starters, and throw in configuration, networking and security for another. No one is arguing that the job of introspecting a running production server and extracting the details necessary to perfectly replicate it in a subtly distinct environment are trivial. The question is whether or not it can be done effectively, at scale, at a reasonable expense for provider and customer, in fewer than 17 steps (not to single out that author, of course).
My bet is that workload migration is absolutely a solvable problem. Further, I’d argue that the economic incentive to solve it is significant: the difference between an addressable market of new workloads and one that can also snapshot and integrate existing infrastructure is potentially massive.
Just one request: please, when you’re migrating the workloads, back up the data first. I can’t take any more of these nonsense “The Danger Sidekick Outage Proves Cloud Computing is Dangerous” headlines – I just can’t.
