When you listen to advocates of cloud computing, you almost always hear that cloud computing is less expensive and more flexible than traditional client-server models. In some ways, that’s true, but the answer isn’t nearly as clear-cut as you might think.
Take pricing. Cloud models are certainly less expensive to get started in than a model where you have to build your own data center and buy your own equipment. So that certainly makes sense for startups and for big new projects.
However, for organizations that already have a data center and enough scale so they can manage it efficiently, and commit to best practices, the answer can be more complicated. As I discussed in my last post, public cloud pricing is often very transparent but very complex, with each little part of the application coming with its own price tag. The answer will vary by application, but I do know a number of big enterprise CIOs who think delivering services via a private cloud can actually be less expensive. Whether that’s true in the long run is an interesting question, but when you already have an investment in equipment, data centers, and personnel, using things you are already paying for is often more cost effective than buying anything new.
The Infrastructure-as-a-Service (IaaS) companies have been lowering prices at quite a rapid pace in recent months with Amazon, Google, and Microsoft all lowering prices and often matching each other. Storage prices in particular have declined rapidly.
In part, that’s because the typical IaaS services—compute, storage, data transfers, and in some cases database—are pretty much commodities. It’s very easy to compare how much you pay per terabyte of storage and not very hard to imagine moving from one service to another.
That’s not the case with Software-as-a-Service (SaaS). These are applications that usually require configuration, often become part of company business processes, and may well require at least some training. Moving from one to another isn’t easy.
So while the established SaaS vendors haven’t been raising prices, they haven’t really been lowering them either. In fact, while they all talk about “paying as you go,” most seem to be moving toward annual per-user contracts (away from monthly pricing). In the end, it’s not clear how much different this really will be from paying maintenance contracts in the traditional model. The big difference is you don’t own the software license, so you can’t choose to go “off maintenance” or pick a third-party provider.
But of course, you don’t own any hardware in the SaaS model, so that’s a real cost benefit, unless of course, you already have enough capacity for the application. And such applications may well require less time to manage, and that, too, can save costs.
Flexibility is another story. Cloud vendors all talk about how you can scale up or down as you need to. But again with SaaS vendors moving toward monthly contracts, they aren’t becoming that much more flexible than the annual maintenance renewal that accompanies traditional software.
In some ways, SaaS was designed to be very flexible. Most of the offerings have extensive APIs, giving you the ability to choose plug-ins to extend the application or to connect the application with other tools you are using. Of course, lots of traditional software has APIs as well, but I do find that the SaaS vendors have often gone farther.
But of course, the promise of more flexibility certainly doesn’t apply to the core software itself. You can always configure the software and use various plug-ins, but you can’t really do the kind of customization you could do with on-premises client-server versions. And when the vendor upgrades the software, you are automatically upgraded—it’s usually not possible to be running on an old version. This has its pros and cons, but it is hard to consider it more flexible.
So while cloud advocates talk about price and flexibility benefits, it’s important to keep in mind that it’s both a matter of perspective and highly dependent on the applications you are running. Cloud services can be much less expensive than building a new data center, but in daily operations, they may not be that much less than an on-premises offering. In some ways, most cloud offerings are quite flexible, but in others, it is quite static—everyone has to run the exact same version.
(article by M.J.Miller)