OK, let’s say you’re a CIO who’s promoted hybrid cloud computing in your company. Then along came all these news stories that call into question the whole notion of cloud economies. Do you send some covert IT team to block the news from the CFO’s computer, or do you deal with it? Hopefully, the latter.
I’ve examined audits of over four-dozen cloud projects, and the good news is that most cloud applications make the business case. The bad news is that a lot, a worrisome lot, don’t. If you want yours to succeed, there are some strategies that will help, in the form of five “forgets” and four “remembers”.
The first strategy is to forget cloud economy of scale as the cloud justification. Enterprises often think that hybrid cloud providers with enormous data centers will have much better economies of scale than the enterprises’ own data centers can deliver. Not true; economy of scale doesn’t just keep improving as data centers get larger, it reaches a plateau. Many corporate data centers can come within a few percent of cloud economies, a difference too small to cover cloud provider profit margins. So, if you think cloud is always cheaper and launch projects on that basis, you’re in trouble.
The second strategy is to forget “moving” applications to the cloud. The biggest benefits of cloud computing are scalability and resilience, but applications written for the data center can’t normally exploit those “cloud-native” benefits. Not only that, when applications are “moved” to the cloud, the work has to cross between cloud and data center, and these crossings incur additional charges. If a database is hosted in the cloud and accessed from the premises, or vice versa, the result can be a cost explosion. You have to redesign applications to realize key cloud benefits.
But forget rewriting core business applications for the cloud unless you’ve accumulated a lot of cloud experience. This is the type of project that enterprises say is most likely to fail, in fact. If you use third-party software for your core applications you may be able to get a version designed for the cloud, perhaps from a different source, but even there, you’ll need to validate vendor claims or face a big failure risk. A third of enterprises said their software vendors misrepresented the “cloud-native” state of their products.
Next, forget the idea that all you need for cloud success is better monitoring or “observability” or more “automation” or perhaps one special new tool. Enterprises tell me that cloud business-case failures aren’t subtle; there’s almost always some glaring issue with plan or execution, and nothing short of a major project do-over is going to fix it. Nobody wants to admit that, but throwing more money at a bad plan is only going to make things worse, and more costly. You have to get your cloud plan right from the start, or grit your teeth and start over.
For our last “forget”, forget fancy cloud features like “serverless” and cloud-provider web services unless you have proved they’re essential. There are many kinds of cloud services and many optional features that will ease developer burden. They get a lot of favorable attention, but most of them are designed for special situations, and all of them will end up costing you more unless you fit into one of those situations. Most business applications won’t, and if yours needs something special, you’ll find out in the prototyping process I mention below.
Despite these points, there’s one class of cloud application that makes its business case so often it dominates cloud success stories. That’s the “application front-end”, where cloud components are used to create a layer between users and legacy data center applications, to add support for mobile devices, improve the user interface, and scale website processes. But remember that the network and network security is a key part of your front-end application. Users will have to get to the front-end cloud piece from the Internet and sometimes also from your VPN, and traffic will have to move between cloud and data center. There are a lot of new attack surfaces that could open, so be sure they’re taken care of. The cloud is no more intrinsically secure than the data center, and if you don’t take care you can make it less secure.
The biggest challenge the cloud poses for your CFO is that it breaks the normal fixed-cost model of the data center. It’s therefore important to remember to do very careful cost simulation for cloud-native applications, over all the possible scaling and resilience scenarios you want to support. This is especially critical where your applications scale under load because scalability improves performance for users by increasing cloud costs. You’ll need to understand how much you’re willing to pay, and write scalability limiters into your applications to ensure you don’t go outside your acceptable cost range.
It’s surprising that this point needs to be made at all, but it’s critical to remember to fully test your cloud application at every point, before you’ve made any big commitment. Start by prototyping all your ideas, and refine and test until you’re happy. Testing doesn’t mean just functional testing either, it means testing the assumptions that your cloud business case depends upon. Is the performance what was expected, the cost in line through the range of conditions? Are the operations and support impact what you expected? Most companies don’t have a lot of cloud application experience, and it’s easy forget that component scalability easily translates into scalable spending.
For our final strategy, remember to qualify your development and operations staff fully in cloud technology before you even start to evaluate cloud usage. Of the 64 cloud project failures I’ve looked at over the last 18 months, 62 were due in part to lack of qualified enterprise staff. In most cases, the problem with a skill shortage is that it contributes to bad planning. There are so many cloud misconceptions out there that one of them will bite you for sure, unless you have people that can help you find the truth.
What is the truth? That the cloud is another hosting choice, and that no hosting choice is right for all situations. That for what the cloud is good for, it’s the best thing around, and what it’s bad for, it can be terrible. Learn what the cloud is good for, and it will be good for you.
Copyright © 2021 IDG Communications, Inc.