When Amazon is your biggest competitor, you must have a top-notch engineering department and a laser-sharp focus on your online business if you want to stay relevant in the tough world of retail sales.

For Lakshmi Sharma, vice president of cloud and compute at Target, that means figuring out how the retailer can use software to achieve efficiencies across private cloud, public cloud, and the data center. Sharma, a software-defined networking (SDN) expert and veteran of firms like Brocade, Cisco, and Rift.io, recently talked with SDxCentral Editor-in-Chief Sue Marek about her top IT priorities at Target. This interview has been edited for clarity.

How do you transition from legacy applications and infrastructure to a more virtualized environment that is state-of-the-art? 

The good news is that our current leadership comes from a background that has made this transition from legacy, to the new world of virtualization and cloud. They have experience in this and know what to do to grow the business.

One message from the leadership team was that we needed to consolidate in terms of vendors. There were a lot of contractors working from different agencies, and there were a lot of partners that were offering similar services.

We consolidated those vendors and that gave us an inventory of what was useful and what we could retire. That is where we started.

We reviewed all our applications and consolidated them or retired those that were no longer necessary.

Give me an example of an app that you retired that no longer was necessary?

People in IT are often afraid of migrating legacy apps that are running on the mainframe. We made an effort to look at what was running on the mainframe, and then we identified what wasn’t being used. That was a huge benefit.

Secondly there were demands in different parts of the company, and there were projects in which each team would build a solution to help the customer.

While that made the company faster, each team ended up choosing their own technology and execution. There was a lot of virtualization already in place because of that. But it created a lot of different versions of operating systems and different versions of hypervisors, and job tools, and tools for building the code.

The first thing we did was consolidate it and move from a services level view to a product level view. Then we helped educate and train the workforce from the leadership down to the DevOps teams. That helped us consolidate the operating systems and software and still maintain the speed and agility that is required.

What percentage of your apps are in the public cloud vs. the private cloud, and how do you decide which apps go where?

We continue to evaluate what is going in the public cloud and what goes in the private cloud. The general philosophy is that anything that is bursty in nature, or if you need to instantiate something that is required for that hour or two, you will want to use the public cloud and make use of that elasticity of the public cloud.

Also applications that are guest-facing that require diversity in terms of geographic locations and latency. In those scenarios, the public cloud will have reach in different data centers and be helpful.

What we want on-premises is anything that is being regulated. For companies like us that are in the retail industry, we would prefer to keep them on-prem. This is mostly for monitoring. Public cloud providers offer security and things, but in certain cases the cost of security that you have to pay for this data that must go over the network to the cloud, may not be justified.

What are some of the challenges working with cloud providers?  Are they really agnostic?

That’s a very interesting question. I would not call them cloud agnostic. Applications that are cloud enabled means that I should write my application in a way that could use some set of APIs, and based upon those I could be elastic and horizontally scale my applications. I could attach to storage based upon the logic of my apps.

What I have learned is that each cloud has its own fabric or underlay — each of them having their own versions of network fabric or compute fabric or storage fabric.

All the cloud providers need you to work with a third party to be able to deploy an application in an agnostic way.

You can’t move an application from one cloud to another without writing logic around the underlay. And you can go to a third party to write that abstraction layer. But thinking that you can move an application from one cloud to another without investing into anything other than an application is not real. It is not seamless when you move from one cloud to another.

I feel like we are moving from mainframes that were boxes to mainframes that are the cloud.

What about the cost of moving from one cloud provider to another?

There is no apples-to-apples comparison of the costs. If I have an application and I’m using one cloud offering, and I look at using another one, there is no way to compare one to another. There is no calculator that goes from one to another. I believe that is how they are designed internally. The cost structure looks very different.

When you are an enterprise trying to use all these capabilities, like monitoring and data storage, there is no way to compare cost structure because this financial modeling requires a lot of effort.

You’ve been at Target for about five months now. What are your top priorities?

My first priority is to achieve efficiencies across private and public cloud and the data centers. That’s my IT goal — to achieve efficiencies around resources, to achieve efficiencies on infrastructure and software, and all the technology pieces we have.

The results should be to improve guest experience.  How would these efficiencies be used to improve the guest experience?

And my second priority is how do I support the digital growth? Everything that we do in IT and software development should be done to support the digital growth of the company. We have grown our online business more than 200 percent in three years. How do I build that infrastructure so that the growth continues?  And how do I make a smoother experience between the digital world and the store world?

How do I enable my infrastructure and my team so we deliver a ubiquitous experience across digital, across distribution centers, across stores, and the team member to the guest experience?

My third priority is how do I use my stores to support that growth in digital? We need to create a ubiquitous experience for developers across cloud and data centers and give a unique and seamless experience to the guests and the team members across the digital business and the stores.