All clouds are fast and elastic, right? Well, it depends on many factors, but I can say for certain that the way you use cloud determines performance more than the cloud itself.
Cloud computing is really a new way to consume resources. The resources themselves are not much different than the resources you’ve consumed for years in private data centers. What’s different about the cloud is your ability to spin up any number of resources, as needed. Once you’re done using them, you can spin them down.
This is why public clouds are typically elastic, although this does not ensure both performance and scalability. To get both of those attributes, the key to success is more a matter of application design than it is the cloud platform itself.
Poorly designed applications perform poorly no matter where you run them. Yet many enterprises try to take these applications, move them to the cloud, and hope that all is magically fixed because they’re now on the cloud.
The reality? Moving to public cloud platforms can actually make poorly designed apps run worse than before. In many cases, problem applications may consume more resources — and cost you more money on a public cloud — than they did in your data center. And the latency in communicating between the cloud and your users over the open Internet could actually slow things down even more.
Don’t get me wrong: Public clouds can provide better performance than on-premises systems. But that happens when they’re designed well, including addressing the reality of latency.
If you’re porting applications to the cloud, perform an evaluation that defines how they can take advantage of the native features of the public cloud provider you’re looking to use. That’s how you get the best performance in the cloud.
Fortunately, enterprises are starting to learn this lesson. How do I know? Because I’ve noticed a surge in the number of job postings for cloud application architects. The use of that talent means companies are no longer just lifting and shifting to the cloud, but are now understanding that some re-engineering should occur to get to the desired level of performance. Without that evaluation process, you’ll have a suboptimal application on the cloud — and you’ll lose some, or even most, of the cloud’s key value.