AI Networking with Broadcom's VeloCloud: Innovations with VeloRAIN and Beyond

Discover the transformation of VeloCloud from its startup days to becoming a significant player in the AI networking space under Broadcom's umbrella. In this episode, we have Sanjay Uppal, VP & GM, Software-Defined Edge Division from VeloCloud by Broadcom, who reveals the secrets behind their latest groundbreaking architecture, Velo RAIN, designed specifically to tackle AI workloads beyond the confines of data centers. Learn how Velo RAIN is redefining AI networking by effectively addressing challenges such as user experience, security, and cost. We delve into compelling real-world applications, such as enhancing new wireless underlays like fixed wireless access and bolstering secure communication between AI agents within enterprises.

=Sanjay shares how VeloCloud is enhancing the user experience with real-time data and seamless network performance. The competitive landscape is heating up, and VeloCloud is right at the forefront, ensuring that smart technology delivers on its promise. Whether you're curious about the future of AI networking or the next big thing in retail tech this episode offers a treasure trove of insights and innovation

Transcript

Speaker 1 0:01

And I'm here at VMware Explorer with one of my favorites, Sanjay from VeloCloud by Broadcom. Sanjay, how are you?

Speaker 2 0:08

I'm doing very well, Evan. So nice to see you.

Speaker 1 0:10

Good to see you. It's been a minute and you have a lot of news and updates for us, and I'm super excited to dive in. For those who may not be familiar with VeloCloud, tell us about the journey from VMware to Broadcom and maybe some of the interesting milestones along the way. I've been watching you and your team for many years.

Speaker 2 0:31

Yeah, no, it's been really exciting. So this is Velo 3.0. It's the third incarnation. First, of course, was Velo 1.0, when we started this as a company three guys in a garage with no money, and you know, we built this technology that solved the problem for enterprises in their wide area network to make it simple to get up to the cloud. And then, lo and behold, much to our surprise and, of course, delight, enterprises really started buying this in boatloads and then it became a category called st van. In fact, we started it even before there was an SD-WAN. Then we got acquired by VMware and, of course, vmware, being very much on the compute side, spent a few years there, learned about what it is to get access to the data center. And then along comes Broadcom and buys VMware, and now Broadcom, by the very name, is Broadband Communications.

Speaker 2 1:21

So, Broadcom is right in the center of networking. So broadband communications, so Broadcom is right in the center of networking. So we become extremely relevant now as one of the 26 divisions of Broadcom really focused on expanding VeloCloud even beyond what we're talking about in terms of SD-WAN because of all this new advent of AI workloads. So AI networking is where we are going with Velo 3.0 and its third incarnation as an independent division of Broadcom.

Speaker 1 1:49

Well, it's exciting times and the big news here, as everywhere, is integrating AI and ML into your platform with a completely new architecture. That's pretty disruptive. Maybe talk about that, because that's the headline today and this week.

Speaker 2 2:03

That's definitely the headline. So we're introducing Velo RAIN and RAIN stands for robust AI networking. About that, because that's the headline today and this week. That's definitely the headline. So we're introducing VeloRAIN and RAIN stands for robust AI networking, not to do with all the RAIN that we've been seeing here in.

Speaker 2 2:12

Barcelona, and it is, as you say, an entirely new and novel architecture for AI networking. Now, of course, there's AI networking inside the data center. I'm talking about everything that's outside. So what happens when enterprises have AI workloads that need to be accessed and those AI workloads can be in the public cloud, they can be in the private cloud or they can be at the edge, and what we're noticing is that these AI workloads behave very differently than the workloads we're used to today, and that's the genesis of this new architecture. The new architecture is needed because the applications and the workloads behave differently on the network and if you don't react to it, it's going to have a pretty poor impact on the user experience, on security, on cost, on performance, and that's why this new architecture called Bellow Rain, and I love your customers and what they're doing with the platform.

Speaker 1 3:08

You always have so many interesting use cases and application scenarios. Do you care to call out one or two that might be of particular interest this week?

Speaker 2 3:17

Yeah, I think that you know.

Speaker 2 3:18

We look at telecom operators and telecom operators have gone out and you know they're deploying new underlays all the time. The most recent ones, especially in the US and in some countries outside in Europe and Asia as well, is fixed wireless access. Right Now, fixed wireless access and its brethren, which is satellite connectivity, these are, you know, faster growing than, let's say, broadband. In fact broadband is not growing that fast. Fiber of, of course, is a very good symmetric technology, so that continues to grow.

Speaker 2 3:48

So a use case really is how do we optimize the applications that are running on top on these new kinds of wireless underlays?

Speaker 2 3:56

And the reason why this is an important problem to solve is because the wireless underlays actually tend to vary a lot.

Speaker 2 4:02

So bandwidth, packet loss, jitter and latency what we call channel, the channel itself varies. So if we can use AI and ML to estimate how the channel is going to behave from past behavior and from training, then we can figure out how to optimize the performance of specific applications that are running on top of enterprises' SD-WAN networks. So this use case we are very excited about go out there and get it done for telecom operators. Now another use case, much more on the enterprise side, and I won't name any specific company because we don't have permission here, but there are enterprise companies, as you've noticed, agentic AI. Agentic AI is becoming the way that new enterprise applications are going to be deployed and, in very simple terms, you as a human being will ask a question and that question will then go to, let's say, a supervisory agent, but that supervisory agent will break that question down into its components and will ask other agents, who then will be backed up by an SLM or an LLM.

Speaker 2 5:06

And then the answers will come together and you get the answer back. So what that really means for us in the networking space is that these agents then talk to each other in a peer-to-peer manner, and our responsibility as VeloCloud, with this VeloRain architecture, is to set up a set of segments, in fact, one meshed segment, so that the security and the performance for these agents to talk to one another is done in the best possible way with agentic AI.

Speaker 2 5:34

So, two use cases one on the fixed wireless access side, other with agentic AI for enterprises, and both of those I think you know. We're already seeing that the take up rate for these is becoming really, really strong. So we're at the right time, I think, with the right solution.

Speaker 1 5:50

And your go to market is also evolving and you've announced the Titan Partner Program to support MSPs. You have thousands of MSP partners. Maybe talk about the go to market and how that's changing dramatically for service providers and enterprises?

Speaker 2 6:05

Yeah, so managed service providers is really the heart of our go-to-market. Of course, we do some direct as well, but MSPs and they could be MSPs as telcos or they could simply be MSPs as overlay MSP providers. For both of those, the Titan program simplifies how they deal with Broadcom and the simplification. Just one aspect of the simplification is they have licensed portability now. So Veloclart has always been a SaaS offering, and now what they can do with the SaaS offering is, if you're an MSP, you can buy licenses for a particular end customer that you have, but then you can also take these licenses and then apply them. If the first customer is not using them or they've upgraded into something else, you can take this license and apply it to a different end customer that you have.

Speaker 2 6:51

You don't need to necessarily come back to us and ask for permission. You needed to do that before. We needed to keep track of your end customers. We don't need to do that before. So I think this really elevates the game. This makes it much simpler for MSPs to deal with us. We're also, because of this, being able to reduce the amount of transactions we do with the MSPs, so they can bucketize their licenses that they need to purchase into chunks, whether it's monthly or quarterly, and so the number of transactions that they have to deal with and that we have to deal with dramatically go down Again makes things very simple. Under Broadcom, simplicity, transparency really rules the roost, and so that's what we're adopting for our Titan program as well.

Speaker 1 7:33

That's exciting and I'd love to get into hardware and SKUs and speed and feeds. We'll save the audience from that. But you do have new edge appliances that sound exciting and quite differentiated. Care to touch on some of those and the opportunities for customers.

Speaker 2 7:48

Yeah. So let me go first to the low end of the spectrum. We have the 710 5G, which is a device that does fixed wireless access. It does it really well because you can set up much better performance compared to regular fixed wireless access. But the really interesting part of that device is also you can converge it with a wired infrastructure so you can think of fixed wireless access on one side, fiber on the other, and bring in true convergence. It's not just convergence on your bill, it's actually packet-by-packet convergence, so your wireless and your fiber can look like one network. And it doesn't have to be fiber, it can be broadband or any other network, it can look like one network for a specific application and look at two networks for a different application.

Speaker 2 8:29

So it's an application-by-application convergence on the 710 5G. Then we have our two high-end platforms that we've introduced, the 4100 and the 5100. The 4100 takes the speeds up to 30 gigabits with a consequent increase in number of tunnels, and then the 5100 takes it to 100 gigabits. So now you can think of data center to data center traffic, very large hubs, and both of these we are calling as AI ready. Ai ready because not just our algorithms run with AI ML, the new versions of them, but also you can host SLMs on these with the right hardware support in them.

Speaker 2 9:09

So 4,100 and 5,100 high-end platforms for large enterprises AI ready.

Speaker 1 9:15

That's fantastic, and you've been an innovator in security for some time and you're only doubling down on security investment innovation, hardening. Talk about some of the challenges out there that your customers are seeing from security, end-to-end security.

Speaker 2 9:28

Yeah.

Speaker 2 9:29

So the first one really is what segmentation solves which is this data that is coming in from the endpoints of an enterprise because of AI, because data is used. Once the foundational models are trained, they need to be kept fine-tuned and for that you need to use something like RAGG. But RAG means that you have data coming in from the endpoints of an enterprise that needs to feed these SLMs or LLMs. So what you do is you use segmentation to put that data in a particular segment so that you can restrict the attack surface, and we have many segments that we can support at an overlay layer, so these are virtual segments that run on the first, on the physical interlay. So you have segmentation. Then, on the endpoint, at the edge, we have introduced a robust edge firewall. We've got IDS, ips, url filtering.

Speaker 2 10:19

So, without cracking open the keys of the transaction, we can do security to prevent threats from coming in and attacking the edge itself and everything sitting behind it. Then that edge connects through segments to the data center through the cloud, and what we have in the cloud is our SSE offering. So think of that as a place where we can. We get the keys, we open up the transaction and then we can look in. We can block attacks over there as well. We can do content access security brokering to figure out what malicious content there might be and then also block data leaks. So, because of AI, a lot of information is being fed up into the network and what we need to do is block malicious traffic. We also need to block users who are leaking away enterprise data. We also need to block users who are leaking away enterprise data. So all of this in our security architecture that is akin to the networking as a service that we're doing with Velo.

Speaker 1 11:14

That's extraordinary. So as part of Velo 3.0, there's lots of synergies within other divisions within Broadcom on stage here at Explore. Are there some you're more excited about or what are the game changers there for customers?

Speaker 2 11:28

Yeah, so first of course we are part of, you know, we were part of the VMware family. So, for us, as I was describing on stage. You know, if you have VCF in the data center, whether it's public cloud or private cloud you can use Velo to link that up into anything that you have at the edge, including our edge compute stack. So Velo becomes the connectivity fabric for your private cloud as well as public cloud infrastructure.

Speaker 2 11:52

So, that's the first one. Then, if you look at the classic Broadcom divisions, of course we have a very close partnership with Symantec. So our VeloCloud SASE underneath is secured by Symantec.

Speaker 2 12:03

But if you're an enterprise, customer or a telco. You come to Velo and we broker, of course, the internals of how we use Symantec and we package it together, we integrate it with VeloCloud so that there's only one VeloCloud SASE out there, so we have a good partnership with them. Then there's actually an immense amount of really cool technology that is being built business-relevant technology by other divisions of Broadcom. So if you look at what's going on with the broadband video folks or what's going on inside the data centers with Ethernet, these are all potential opportunities for us to work with other divisions of Broadcom to incorporate them into VeloCloud. In terms of whether you're looking at machine learning, whether you're looking at faster traffic, all of these are opportunities for us. I don't have anything specific that I can mention today, but I can tell you that it's pretty exciting to be part of this journey along with these other groups at Broadcom.

Speaker 1 12:58

Well, watch this space for more updates and innovation. Any thoughts on what else you're looking forward to as we head into the busy year end and beyond? What's on your personal and professional radar?

Speaker 2 13:08

So let me put on my smart glasses.

Speaker 1 13:09

I saw them on stage.

Enhancing User Experience With Smart Glasses

Speaker 2 13:11

Very very cool glasses, and the reason I put them on is because this is one example of how the traffic is changing so rapidly A 100 times more information going upstream through these two cameras that you see here, as coming downstream through the speakers that are at the back of this pair of smart glasses. I'm very excited about these. The reason I'm excited is because we can improve the performance of them, because what you need as you walk into a store or you're going about your day if you are live streaming from these, you need as you walk into a store or you know you're going about your day if you are live streaming from these, you need real-time feedback. That real-time feedback needs a network that works. It is secure, highly performant, and so we are going to improve the quality of experience for users with these glasses using VeloCloud, and I'm very excited about that.

Speaker 1 13:58

It's exciting and not just meta. Apple's jumping into the game, evidently with their smart glasses, and all the cool kids will be wearing them. I include us in the cool kids category. But thanks so much for joining and congratulations on another successful Explorer.

Speaker 2 14:12

Thank you. Thank you, evan, and appreciate the chat. Take care, bye-bye.