Lori and I were talking shop over the weekend, and one of the things a co-worker had said on an email thread triggered an entire series of observations from me. Not being one to keep my mouth shut, I thought I’d share them with you and lead you to my conclusions, so that you can decide if they’re accurate or not.
The presumption of the hype cycle is that you need and will adopt cloud computing en-masse. This is an interesting presumption, but the security, economics, and Service Level Agreements (SLAs) will have to be there long-term for it to stand any chance of being true. Of the three, I see the largest being SLAs. Where the line is drawn between the cloud provider’s responsibilities and yours will be the weak point. Every IT shop I know that has gone with outsourcing or SaaS providers has gotten into the problem of “who’s problem is it anyway?” where the vendor points back to IT, IT staff thinks it is the vendor’s problem, and valuable time is lost while it all gets sorted out. In the end, someone is right and someone is wrong, but that’s not nearly as important as the fact that systems were down while you figured it out and it took a lot longer to figure out than it would have if one person or group were responsible for the whole system. With internal cloud, one group or person essentially is – the IT department that would be responsible anyway, so I’m sticking to a discussion of public cloud in this blog.
You see this sometimes between groups within an IT shop even. A NAS device becomes inaccessible, and the network/storage admins point fingers at each other. It’s not nearly as common as the external vendor scenario, but it happens. In the public cloud you will have at least three parties that will share in responsibility for your application’s availability – the Internet provider (four parties in the highly likely event that you and your cloud provider do not share an Internet access vendor), your staff, and the cloud provider. There is lots of room there for things to get out of hand, and you’ll have to negotiate your contract in such a way that your vendor has to work with you until an outage is pinned down and the cause definitively known.
The next most important is security. Seriously, those who claim cloud security “is not a problem” are not betting their careers on it. IT Managers are, and that fact alone should make you tread with caution. It’s not that cloud is inherently insecure (I’ve weighed in on this topic with my thoughts previously), it is that you need to prove it isn’t, or you may find yourself fired for a huge data leakage, just the type of firing that makes you spoiled goods and unlikely to land an equivalent job elsewhere. So be wary, if you go with public cloud, make the vendor prove they’re secure enough for your organization’s level of risk aversion.
And finally, the financials. The rule of the hype cycle is the inverse of “What goes up must come down”. Let’s call it Lori’s Law - “What the hype cycle drives down in price must come back up”. Sure, early cloud providers are going to cut to the bone to get your business, but that’s not the best way to run a company, and eventually the cost of cloud computing will go back up – either through addition of essential for-fee services, or increases in base rates. While many things in the Internet are artificially depressed in price because a few clicks of a mouse and a few thousand dollars can create competitors, the horsepower and marketing required to sell something like cloud computing put it out of that category. A few big vendors will end up running the show – not necessarily vendors that are big today, just a few big vendors – and prices will raise over time. Get price lock-in with your contract, and make sure you have an exit strategy before you go moving critical parts of your business out there.
Still, the pundits assure you that you need cloud computing. If you have a single datacenter you need it for redundancy, if you have multiple datacenters, you need it for data/load sharing, and if you have no datacenter, why in the world would you build one when you can do it all in the cloud! (imagine hand waving and screaming here)
That’s not real close to reality. Their arguments are valid, but they ignore the expense of moving things to the cloud, the changes in staffing, architecture, and deployment models that must occur. In short, they ignore anything that is an inconvenient truth. Otherwise we’d call them something other than pundits. And the sage advisors that tackle these issues head-on and are trying to help you understand both whether you should be going to the cloud and how to get there in terms relevant to most enterprises today? They’re not in this group I’m referencing, they’re helping you figure it all out.
But all of this lead me to question one of the fundamental assumptions of Cloud Computing: That you, the enterprise, need public cloud to be competitive, adaptive, effective.
What if you don’t? Specifically, what if you could get all of the benefits out of the cloud without ever actually using public cloud?
Software As A Service (SaaS) has been around a good long while now, and some companies have managed to be successful at it in the face of many of the same barriers that Cloud faces today. Cloud has the benefit that it is new enough that you all will be out there testing it out over the next couple of years, and your decisions about what to use it for will greatly impact the cloud vendor landscape. SaaS never had that – because it was Software as a Service, but was sold as a specific application or suite of applications, the market as a whole didn’t live or die in the same manner. Salesforce.com was smashingly successful while other similar offerings failed. I’ve pointed this out before, but coming at it from a different angle.
And that is the completion of my thought. Cloud could become the deployment model of choice for SaaS providers – people who need unlimited potential connections, ultra-high-performance, and truly unlimited growth potential with pay-as-you-grow requirements – and you, as their customer, would reap the benefits. It is far easier to say “let’s purchase application X as a Service” than it is to say “Let’s move our entire infrastructure to the cloud”.
And what if that’s the correct answer? What if email and Azure and Google Apps, the whole pile of other applications running on the cloud are the real wave of cloud computing, and we’re all just wasting our time talking about massive public cloud migrations of enterprises?
It’s worth considering anyway. At least then there is one organization reporting for application availability and stability, not four. And at least then you’re not throwing the baby out with the bath water, as some in the cloud digerati community have been suggesting you must do for years.
I’ve been around the hype cycle a good long while, and this rings true for me. It makes more sense that we as an industry would make an evolutionary change than a revolutionary change at a time when Virtualization of the datacenter isn’t even completed yet. Internal cloud is evolutionary from Virtualization, external cloud is revolutionary from a “way we work” perspective. This skirts the line well.
Related Articles And Blogs