Denys Rudyi - Fotolia
SAN DIEGO -- Every August, dozens of Gartner analysts put their research, expertise and predictions on screen and on stage for several hundred IT professionals who gather for the annual Gartner Catalyst Conference.
Gartner research vice president Kyle Hilgendorf, who specializes in cloud and the Internet of Things (IoT), sat down with SearchCloudApplications to discuss the wisdom of adopting a cloud-first strategy, porting legacy applications into the cloud and the critically important -- yet often overlooked -- step of estimating monthly cloud expenses. In upcoming installments, Hilgendorf examines issues related to maintaining control over an incoming potential torrent of IoT data, monetization of that data, the current competitive landscape among the five leading cloud platform providers and the future of application developers.
SearchCloudApplications: You've remarked that a cloud-first strategy does not mean cloud always.
Kyle Hilgendorf: I firmly believe that we're in a transitional state. Cloud computing brings in a lot of benefits that solve problems of agility, scalability and unpredictability. But, we have a huge existing IT application portfolio that was never designed with an understanding of what cloud computing is. So, that type of application will need to continue to run in more traditional environments.
No one can predict the future. A cloud-first strategy means that you put cloud as a priority in your organization. Even for applications that make perfect sense to go to the cloud today, the characteristics or business process may eventually change. At some point, you may decide that cloud is no longer the deployment model. Maybe some technology revolution will supersede cloud. We can probably bet that will happen at some point.
Are legacy CIOs who think in legacy terms an impediment?
KH: Just like in any role in the industry, you're going to have forward-thinking thought leaders, middle-of-the-road thought leaders and some people who are stuck in more of a previous mindset.
What inherent dangers do you see? And what are the risks of failure in simply porting existing applications to the cloud?
KH: Porting existing applications to the cloud is going to run you into a series of problems with cost inefficiencies. If an application is not designed to be dynamic, elastic and scalable, then running it in the cloud in perpetuity for 24 hours a day will not reach the cost model comparison against an on-premise data center. It also could lead to some unfortunate performance or unavailability events within the cloud.
Kyle Hilgendorfresearch vice president at Gartner
Here's a simple example: Consider that if your application in its current or legacy state cannot be designed to be horizontally scalable, and is tethered to a single operating system or single virtual machine, and then placed into a single data center in the cloud provider without taking advantage of the resiliency, availability and scalability designs, you could run yourself into an unavailability event with a single machine. A refactoring or rebuilding of that application would lead you into a parallelized environment in multiple data centers with load balancing.
It's been said that if an application cannot run in a browser, it has no business being migrated to the cloud.
KH: I don't think that's an accurate statement. Web applications are very easily defined use cases in the cloud, but we see a whole variety of applications that might be mobile apps and not browser-based, or even more traditional client-server applications that are being designed for a scalable and more parallelizable world. We have a lot of back-end applications running in the cloud among Gartner's client base that are not Web-enabled whatsoever. They might have a Web front end somewhere that could also be in the cloud, but the back end is scaling the databases, data warehouses and middleware.
One thing that's not clear among platform providers is billing, though the scourge of hidden charges has improved. What do we need to do in terms of cost and usage estimating, and governance to avoid overconsumption that can lead to unexpectedly high monthly bills?
KH: We encourage our clients to go through a cost simulation phase. This is profiling all the characteristics of your application or portfolio that you're considering moving into the cloud, and leveraging either your provider's cost calculator that most make publicly available or opting into a third party -- where input all the parameters about your application. It spits out an hourly, monthly and annual price estimate. That's a good first start. It's not going to be a perfect scenario, because that estimate is only as good as the predictions about that application at that time.
That leaves a lot of room for variability.
Kyle Hilgendorfresearch vice president at Gartner
KH: If the application tends to be more popular or starts to scale out, or throughout the lifecycle of that application you determine for resiliency reasons that you need another replica in another data center, you are potentially doubling your costs. After the cost simulation has happened, let's actually run a point-in-time proof of concept in the cloud and see what those bills look like for the first couple of months. After you get through simulation and profiling, and start to run it in a production environment, there's going to be a continuous phase of optimization and analysis to determine if there's anything else you can do to the application or any configuration changes you can make for cost efficiencies.
In the part two of Hilgendorf's interview, he examines issues related to IoT data collection and monetization, the competition among cloud platform providers, and the skills that developers and architects will need to master moving forward.
Five steps to cloud-first security
Seven challenges in cloud-first strategies