On-Premise versus Cloud: The Better Option

Denise Zabawski, CIO, Nationwide Children's Hospital

The convergence of several factors recently helped us integrate a larger infrastructure presence in “the cloud” into our IT strategy. This year we recruited two nationally recognized Genomics researchers and their teams. At their previous institution, they were utilizing local hardware which was not providing the needed flexibility and performance. Even before they arrived we agreed that we needed a deep dive into available cloud offerings.

Our current Genomics researchers were already doing some projects in the cloud, so we were not complete strangers to the technology. We had also been using IaaS and SaaS solutions for some time. As part of our institution’s strategic plan, we were also planning to build a new data center. We compared the top cloud players, consulted experts and had the vendors come in and show us their wares. We completed total cost of ownership calculations, including on-premise data center costs for one, three, five and 15 years. We compared the cloud vendors with a flexible on-premise solution where we would just pay for the compute and storage we were using. This is the hardware vendor’s way to compete with the on-demand cloud pricing. The on-premise pricing was very compelling and based on the information we have today, had a lower total cost of ownership. However, by eliminating the on-premise option from the data center we were able to reduce the data center footprint considerably, shaving about $10 million from our construction estimate.

We busted the myth in our organization about the cloud being significantly cheaper. In many cases it is not cheaper than on-premise solutions. It may be cheaper if you don’t already have a data center and IT staff. But if you have all of those things, it is going to be more expensive than what you can do on-premise. If I was starting a new firm, I would run it all in the cloud. Since we have existing infrastructure assets and our healthcare business requires certain on-premise solutions, our solution will be a hybrid one.

  ​We believe a hybrid model, one that utilizes the best of both worlds at the best price point, is the best model for our organization  

Ushering the Cloud Era

Even though we are “in the cloud” already, I learned a few things as we look to build a large infrastructure presence. There is a whole new set of marketplaces. You can purchase a plethora of cloud tools to help you manage your cloud assets, do monitoring, bolster your security, optimize performance, etc. You can also purchase services from a broad spectrum of partners from the cloud provider themselves to certified partners, to the guy down the road. You can even resell your own compute and storage if you are not using it! We picked one of the big players in the market and are designing a proof of concept for the particular type of compute and storage we need to insure there are no bottlenecks that impact our genomic analysis workload. We are working with a broker to get direct, high speed connectivity to that resource. We agreed that our hybrid strategy is one where we use the cloud as an extension of our data center, making smart choices about what fits best on site and what can or should go in the cloud. We hope to be live by April 2017 with our first production instances.

Taking the Next Step

The next steps include activities very similar to those needed for on-premise infrastructure. We are building a gold server image to use on all the instances we spin up in the cloud. We are designing our roles and security. We realize we also need to create policies and procedures as well as revise some of our existing policies to take into account our new infrastructure presence in the cloud. We will be creating a server request process as well as a storage request process.

Specific to this cloud project we are setting up a direct fiber connection to a cloud services broker. Once we drop a circuit to this broker we will be able to have a direct connection to all of the big IaaS cloud providers. We are purchasing a 1GB pipe to start which will be expandable to 10GB when the time comes. We are estimating the 1GB will take us through the first 12 months after which we will probably need to increase to 10GB. We are also discussing what staffing will be needed to handle the accounting and finance component of managing the cloud instances. Without careful management of cloud resources, we could quickly see budget over runs.

Over the last few months as we’ve been working on this IaaS project we have also signed up for online productivity suites in the cloud with multiple vendors. We’ve decided to maintain a diversified cloud portfolio using vendors that provide the best services in a particular category. Our long term strategy is to utilize cloud services as an extension to our data center and service offerings. Our goal is that these services will be seamless with no impact to the customer other than the positive impact of more features and the ability to do more with mobile devices.

Cloud services are now integrated into our IT strategy allowing us to be more agile and innovative while adding a wider variety of services to our customers. There are definitely advantages but there is also increased cost. I’ve heard many of my peers talk about their goal of moving their entire data center to the cloud. We believe a hybrid model, one that utilizes the best of both worlds at the best price point, is the best model for our organization.