In analyst Zeus Kerravala’s Zkast video series, he sits down with tech industry leaders to get their insight. Last month, Zeus had two insightful conversations with executives who are driving VMware’s SASE and edge strategy. The first exchange was with Sanjay Uppal, SVP and GM of the VMware Service Provider and Edge business unit. Next week we’ll bring you the highlights of Zeus’ chat with Craig Connors, VP and GM of the VMware SASE business.
Sanjay and Zeus discussed:
How to define multi-cloud, and the abstraction layer that’s necessary to optimize it
What the modern edge is, the challenges around scaling it, and the top two verticals where edge infrastructure is gaining real traction
Why “Honey, I shrunk the stack” describes the future of edge infrastructure
Watch the video below for the full conversation with Sanjay Uppal, or scroll down for highlights.
Here are highlights from Zeus and Sanjay’s conversation, edited for length and clarity.
Sanjay: My title is a mouthful, but very simply said, I’m the edge guy. The edge could be where 5G is running, it could be in a branch office, an oil rig, a department store, or even in the back of a police car. The infrastructure to improve network resiliency, for security to make the 5G network run better, or to run compute—all of that is edge for VMware.
Zeus: In some ways, at its roots, VMware has always been an edge company and pushed the limits of where compute has been going. The terms edge and multi-cloud get thrown around a lot, and they tend to have different definitions based on the vendor you’re talking to. From your perspective, let’s start with multi-cloud. How do you define what multi-cloud is?
Sanjay: That term has been thrown around a lot in the last 18 to 24 months. For VMware, it’s an outside-in approach. Our customers—telecom operators, large businesses, enterprises—have been telling us that you have to look at the needs of the application. When you drill down, you find that components of that application maybe should run in different locations, different data centers. For example, the visualization may run at one place in the backend, the number crunching may run at a different place, your ML models may run at a third place. This will lead you to an architecture that has different clouds, both public as well as private, that you might want to connect.
Enterprises and telcos run many applications. When you look at resiliency and disaster recovery, and ensuring that all your SLAs are met, it’s going to lead you to a very similar conclusion: If you have the choice of many back-ends and clouds, as long as you can put them together in an automated way and you can get assurance across them, then of course you would much rather have the diversity.
It’s very similar to what happened in the software-defined data center. Before virtualization, the hardware and software used to be plugged together. Once virtualization came in, you disassociated it, and got a layer of abstraction. We’re trying to build this layer of abstraction that sits in front of all different cloud destinations, so that you’re not choosing a hyperscaler or a cloud destination first and then working backwards. You choose the needs of the application first and then working to where it wants to be. The edge is just an extension of that, meaning that when you look at the other end of it, where does the application need to be accessed from? A human being, a laptop, an endpoint, a sensor, an IOT gateway? When you look at it from that perspective, that’s the edge accessing the multi-cloud. So that’s how this whole thing fits together for us.
Multiple clouds and multi-cloud are not the same thing. Multiple clouds means a client accessing different clouds. Multi-cloud is similar to virtualization: Components of apps running in different public clouds, tied together by a common software-based infrastructure.
Zeus: Let’s come back to the edge later. Multi-cloud is different than multiple clouds. Lots of companies use multiple clouds, but if they’re not tied together, it’s just multiple clouds and not true multi-cloud. In fact, consumer-facing apps such as Wayze are not pulling all their information from one source. They’re pulling data such as location and AI from multiple places. Enterprises have been slow to embrace this, understanding that the risk of failure is a lot higher than for consumer apps.
Early on, abstraction and software created allowed VMware to disaggregate the components of hardware. And now multi-cloud allows us to disaggregate software in a way. Instead of having to live with monolithic software that lives in centralized locations, now we have disaggregated software that interacts with multiple locations.
Now, you did toss in that term “edge,” a term that I hear used in many ways: Telco edge, 5G edge, industrial edge, enterprise edge, campus edge Wi-Fi. Is the edge all those things? How should companies think about what edge means to them? Because with cloud it’s simple: You have a centralized hyperscaler or your own private data center. But with edge it seems a lot more nebulous.
Sanjay: You nailed it by saying multiple clouds is different from multi-cloud. At VMware, we extend the digital infrastructure even to public clouds. We have our stack running on the public cloud in products such as VMware Cloud on AWS, Azure VMware Solution. If you have applications that are composite but run on these different clouds, you have the underlying digital infrastructure that ties them together, which is different from accessing multiple clouds on the client side.
It’s very simple. The edge is digital infrastructure that you put across a number of locations close to where the endpoints are producing or consuming data. It’s software-based digital infrastructure, across a number of locations. The physical distribution of this digital infrastructure is critical because it brings a lot of challenges. If you are distributing to 50,000 cell sites or 20,000 retail stores or hundreds of thousands of locations, it’s very different than giving it to six data centers. Then to be close to where the endpoints are producing or consuming data, you have to ask, where does this have to run? You don’t have a lot of infrastructure to run it in many places. So the shrinking of the infrastructure is really critically important.
That’s a simple definition. Then we get into details. Where are these locations? Are they on the enterprise’s premises? That’s the far edge. Are they in between the enterprise’s premises and a telco point or presence? That’s the near edge. For a telco, what’s the provider edge versus the RAN edge? How thick or thin is it? But it all comes from this underlying definition that we think works for everyone.
The main challenge for edge systems? “Honey, I shrunk the stack.”
Zeus: What are some of the big challenges that an abstraction layer can solve as customers go to this multi-cloud, edge-driven world?
Sanjay: One of the first challenges can be summed up like this: “Honey, I shrunk the stack.” We have a customer in the UK that has put our VMware Edge Compute Stack in the back of police cars. You don’t have a whole lot of infrastructure because you have to worry about things like battery consumption and heat. The number one challenge is, have you been able to shrink the stack? In the data center, you have a lot of cores, memory, power. All of that goes away when you’re talking about these kinds of edge locations.
The second challenge is horizontal scale. If you have the same infrastructure that needs to run in 50,000 locations, that’s a lot different than running in six. How do you get all this infrastructure installed and configured in the first place?
The third challenge is one of automation and assurance: How do you keep it upgraded, how do you keep it operating, how do you keep it maintained without having an army of people to take care of it?
These are the main challenges for an abstraction layer. Of course, you could say that we’ll try and do all of this at the software layer, but then every application has to repeat overcoming these challenges, and that’s pretty hard to do. If you put it in the digital infrastructure layer, which is the software layer, and you get enough support from the hardware, then access to the multi-cloud, then that seems to work well at the edge.
VMware is the Switzerland of digital infrastructure. When companies want more than one cloud, and they need to extend multiple clouds all the way to the edge, then VMware is a very good option.
Zeus: The other benefit from my perspective is that you bring a bit of a normalizing function in the areas of policy and security and management. AWS tools are great for AWS, Azure tools are great with Azure, but it’s not like those are going to work cross-cloud. But it seems like, given that you’re the abstraction company, that’s an ideal role for you.
Sanjay: We think of ourselves as the Switzerland of digital infrastructure. When companies know that they don’t want to go with just one approach in the cloud world, and they don’t want to extend that one approach all the way to the edge, then VMware is a very good option because we provide that layer of abstraction. You get choice, flexibility, and automation assurance, from cell sites to retail stores to manufacturing floors.
Your policy point is quite important. In the SD-WAN space, we came up with a mechanism that scaled very well. You’re able to, from an edge location, have a pool-based mechanism to tell you what your profile is. You can ship an anonymized hardware device anywhere, you don’t have to sticker it with the right serial number. It scales because it’s a full mechanism based on policy. And that can be extended into the RAN, to compute, it can be extended in a variety of ways. These horizontal mechanisms are very powerful at the edge, and they will give you access to that multi-cloud world.
Zeus: I find it interesting that your title is VP and GM of service provider and edge. Historically when you think cloud, that’s been owned by three companies. But when you think edge, that opens the door. You can argue that the 5G operators, mobile operators, and telcos are actually in a better position to offer edge services. Now of course, they’re regional, so this complicated multi-vendor role does come to fruition.
Sanjay: I struggled with whether we should just call it the “edge business unit.” But whenever we talk about edge, when you think of the number of locations, a wider network always has a role to play. That wider network is increasingly wireless. I would not have predicted that in 2022 we’d be seeing such a flurry of new wireless options. It really drives the need for an intelligence layer that runs on top, which is what we are providing. And that’s why when you say “edge,” you have to talk about the network underneath and who provides the network—it’s the service provider.
From an architectural standpoint, these go hand in hand. From a business standpoint, what we found through the SD-WAN example is that almost two thirds, sometimes even 70% of the edge business comes through the telecom operator. Of course, we’ll go to market with system integrators and resellers, but really because of the importance of the network, calling our business unit “service provider and edge” is actually somewhat descriptive.
The two areas where edge infrastructure is gaining real traction are retail and the factory floor.
Zeus: It’s accurate. When I think of edge, the historical use cases have been oil and gas and mining, places where you had a hard time getting good connectivity. The rise of all these different wireless options allows us to deploy more edges in more places. What are some of the emerging edge use cases that more and more companies should be thinking about?
Sanjay: There are two verticals where this is really taking off. The first is in retail—for loyalty, loss prevention, to better understand customer access. There are two types of edges being deployed in retail stores. One is to aggregate large amounts of data to understand customer patterns. And the other is vision, to understand how things like automated scanners should be placed in the stores. There are multiple examples, but there are also examples of what people can do to make the retail store the center of a multi-channel retail experience. You need the workloads to be pushed much closer, because it’s too long a trip to go all the way to the public cloud.
The other vertical is digitization of the factory floor. I’ve spent a significant amount of time on factory floors in the last three months or so. It’s fascinating how this physical machinery comes together, but none of it has been virtualized. A lot of it is sneaker-netting, even up to the application layer. If you get it digitized and virtualized and controlled with digital infrastructure, there’s a lot of change going on. At VMware Explore, I had a gentleman on stage with me from a food processing firm, and they were deploying computer vision to increase worker safety, to have better productivity. They’re seeing 30% better productivity, which is very large when you consider the slim margins in some of these industries. Factory floors and retail stores are where I’m seeing a fair amount of demand just in the last 18 months or so.
Zeus: If we do this ZKast next year at the same time, will we look back and say 2023 was the year the edge became real?
Sanjay: 2023 is definitely when you’re going to see things move out of PoC into deployment.
