Some organizations — and that number is rising with each passing month — have discovered that moving everything to the public cloud isn’t all it’s cracked up to be. It can quickly become very expensive, and organizations find they must make unwanted tradeoffs in stability and service.
To be sure, the public cloud offers great benefits and is ideal for smaller test environments, storage and compute in smallish amounts. Amazon Web Services is approaching $10 billion a year in annual revenue; so clearly they, as well as Microsoft’s Azure and Google Cloud Compute, are onto something. And at the highest level, cloud computing has upended the way we view computing infrastructure in today’s hyperconnected world. We need only look at the rapid proliferation of companies delivering “XaaS,” whether X equals infrastructure, platform or software.
But more and more organizations are coming out of the public cloud and turning to private cloud after finding the tradeoffs in economics, stability and service were just too large.
While a growing number of enterprises are pulling back from the public cloud, there is no doubt that the IT stack is getting “cloudified,” whatever path they choose. Among the companies that have opted for private cloud over public are Dropbox and Etsy, for example.
Let’s explore in a bit more detail why more companies are coming out of the public cloud to an on-premises orientation.
First, as your workload requirements stabilize and run 24/7, you end up paying a lot more for the resources in the cloud than if you were to have your own privately owned infrastructure. Second, there are very limited guarantees of performance and stability of the infrastructure with cloud providers, so you end up having varying performance results, unplanned downtime or lack of resources at unexpected intervals. Cloud service providers design their infrastructure to be substantially oversubscribed in terms of capacity, so you can never be assured of quality service and performance unless you pay for dedicated hardware and infrastructure resources. When and if you reach this point, you are far better off building your own on-premises infrastructure.
There’s also the well-known consideration of security to consider when moving to and staying in the public cloud (this is particularly applicable to the healthcare and financial services industries). Although some cloud providers have addressed this issue reasonably well, it’s still a tough decision to place your core enterprise information technology resources in someone else’s hands.
However, the moment you move to a privately owned and managed solution, you lose many of the benefits of the cloud such as the ability to easily and quickly allocate resources as needed such as scaling up and down resources as workloads and demands change. To ameliorate this, you will want to design your on-prem infrastructure as the cloud providers do, creating a private cloud to replace the public cloud you are coming out of. This will give your in-house users a similar experience as they were getting from the public cloud.
And consider big data. When you have small workloads, it makes sense to run them in the cloud, say in a test environment. But as soon as you start to scale or want to move that test into production and your clusters begin to grow, that’s where you’ll start running into issues of performance, stability and exponentially increasing costs.
But there are ways around this. Some companies have adopted a hybrid model to address the issues outlined above. Others have found ways to decouple and disaggregate compute and storage by moving to a software-defined storage solution, among a small number of hardware and platform solutions that can perform this function as well.
Regardless of the path you choose, simply deciding to shift your entire IT stack to the public cloud isn’t necessarily the right answer, from an economic, performance and flexibility perspective. In most things, there are tradeoffs. The question to answer for your own business is how do you embrace your new “cloudified” IT stack in the hyperconnected world and apply it to private cloud installations as well as in the public cloud, using each solution where it makes the most sense.
S.K. Vinod is the VP of product management at DriveScale and is responsible for the company’s product definition and strategy. Prior to joining DriveScale, he co-founded Xsigo Systems in 2004, where he led the product management and business development teams. After Xsigo was acquired by Oracle, Vinod managed the networking and Netra product management teams. Vinod also spent nine years at Sun Microsystems in various product management roles, launching and managing servers, microprocessors and chipsets.