Enterprise Networks Are Embracing Open Options Presented by Cisco - Tech Field Day Podcast Spotlight

Speaker A: AI has created a massive buzz in the networking space. We are now looking at new ways to build networks to optimize for all kinds of different traffic. And that means that there are a lot of options out there for people to choose from, including open networking options. Are you going to make the right choice when you decide to implement open networking? Could you use a good partner whenever you go that route? In this episode brought to you by Cisco, you the future of enterprise networking is open. Welcome to the Tech Field Day podcast, where we bring together a group of IT technical experts to discuss a single idea about key concepts in the industry. This podcast features a variety of perspectives from members of the Tech Field Day Delegate community and is often associated with one of our events. Tech Field Day is a part of the Futurum Group, and this podcast is also published on our sister company's website, TechStrongTV. On this episode, presented by Cisco, we're going to be talking about open networking. But before we get to that, I'd like to take a moment for our guests to introduce themselves, starting with Scott.

Speaker B: Hi, everyone, I'm Scott Robon. I'm the CEO of Solutional and the host of the Total Network Operations podcast.

Speaker C: Hi, my name is Ryan Booth. I, um, am an AI consultant and architect with Blue Ridge Systems, where I mostly focus on applied AI infrastructure.

Speaker D: Hi, I'm Surbhi, and, uh, I'm the director of Product management, um, in the data center AI networking, um, business unit. I'm so excited to be here and talking to all of you.

Speaker A: And of course, I'm Tom Hollingsworth, event lead here at Tech Field Day. Let's jump into the episode. People that are working in the enterprise network today are faced with a lot of choices when it comes to all of the things that they want to do. The networks are becoming more complex and there are more options than ever on how to build them. The question that most people are asking is, should I build what I built before in the hopes that it will continue to do what I want it to do, or should I investigate new ideas, new software? But there's always the chance that you're going to make the wrong move. In this episode of the TechFuel Day podcast, the future of enterprise networks is open. Now. As I alluded to in the intro, enterprises are really focusing on open networking right now. It is a huge topic of conversation specifically because of things like Sonic. It is a network operating system that a lot of companies are offering and a lot of customers are asking about. Operations and engineering teams are starting to put it into production instead of just running it in the lab. I wanted to start off by asking Scott, what is it about open network operating systems that makes them so appealing to the enterprise?

Speaker B: Well, it's really been an interesting evolution. You know I think I first got hit with this six or seven years ago in the disaggregated networking space where this first step of separating the software and the hardware was something that carriers were interested in. That doesn't mean it's the same as open networking. Um, you still have proprietary software but running on white box infrastructure and other hardware platforms. We've also seen the emergence of Sonic from Azure, um, in the cloud space originally. But in the last two years I've seen more and stronger interest in enterprise, um, and U.S. government networks for open networking solutions. So this isn't um, just an experiment anymore. Uh, there are real enterprise and US government customers interested in moving to open networking for their infrastructure.

Speaker A: Ryan, does that track with your experience working with your clients that there's uh, this push to really embrace these open ideas?

Speaker C: Yeah, it seems that um, the conversation like y' all mentioned is definitely coming back up. Um, more and more, um, there was definitely a lot of traction around this five, eight years ago, um, Sonic made a lot of push, added a lot of features, um, that started adding more and more enterprise grade um, functionality. And so yeah, to see this continue as it is now, it makes total sense now.

Speaker A: Serbi, you're on the Cisco side of the fence and obviously Cisco makes a lot of progress products and a lot of software. But what are you hearing from customers? Are they asking for Sonic? Do they want to run Sonic on Cisco hardware?

Speaker D: Yeah, Tom. So um, you know, maybe uh, my answer would be ah, a little bit longer than Scott and Ryan. But uh, please indulge with me. Um, I would like uh, actually to take all of you and our users not just five years back, but maybe more than a decade back. You remember when this whole software defined networking wave and era started. I mean what did those. I mean that wave was also started by hyperscalers, right? I mean the Google and the Azure and you know, of the worlds. And what did they want? Right? They wanted openness and programmability from the operating system or software or asics that are there in the network switches and whatnot. Right? That's exactly what was the whole idea about software defined networking.

Speaker B: So

Speaker D: the concept of openness and programmability, it's not new, right? Hyperscalers were the one driving at that time and they are the one who started driving Sonic as well. So when your team looks at market signals across enterprise infrastructure, um, you're hitting on the exact reality we see on the ground. The sheer velocity of AI has completely broken um, the traditional infrastructure playbooks. Right. So when an organization builds out a massive high performance multi node AI cluster, um, uh, often because they want to keep sensitive workloads out of the public cloud, they are just not trying to figure out scale, they're trying to really scale human efficiency. If you look at it closely, right, what do they want? They want a simplified automated way to build data centers, networks. They need a unified operating model where a network switch can be provisioned and managed exactly like a Linux compute server. Right. And that's, I think what we are seeing is really driving this massive enterprise momentum for Sonic. But uh, I just want to also say a but that let's be real, okay. Open source is not a blanket replacement for um, uh, enterprise grade proprietary stacks. I have to say that. Right. It's an alternative operating model for specific architectural philosophies. Um, so if, if the entire, I mean we are seeing that the enterprises who look for turnkey feature rich environments, you know, which demand entirely different operational structures than fully aggregated DIY Linux setups, um, then we are talking about that they're not talking about. So our philosophy at Cisco in nutshell, is to meet customers exactly where they are on that journey rather than forcing a rigid one size fits all infrastructure strategy on them.

Speaker B: Yeah, just to follow up on that real quick, to your point Serbi, more and more customers in their journey are finding that uh, Sonic might be an open networking approach, might be a viable approach. Again, not 100%, not 75% of networks, but an increasing proportion that it's really coming uh, into focus much um, more greatly now and not just um, from the DIY world, but platform providers like yourselves that are actually providing support for Sonic ecosystem.

Speaker D: Uh, absolutely, absolutely. And that's exactly why as I said, we are going to meet the customers where they are. Um, and that's why Cisco is absolutely one of the biggest contributors in the Sonic open source community as well. And um, we are going to ride that wave

Speaker A: Now. Surbhi, I think one of the things that people are kind of curious about is the fact that Cisco has always had such a huge presence in the data center thanks to the Nexus switching line and specifically within Xoshi. How does this move towards open networking work with the Cisco portfolio? Because I know that there's been a lot of talk about Sonic, but as you just said, some customers maybe aren't Necessarily like they don't necessarily need what sonic offers and NXOs might be a better choice for them. How do you inside of Cisco make that distinction?

Speaker D: Yeah, no, it's an important question. Um, Tom, uh, answer really comes down to the operational empathy for different workflows, right? I mean let me be completely unequivocal right out of the gate, okay? NXOs and uh, ACI are and will continue to be the absolute premier choices for mainstream enterprise networks, okay? The rich feature depth, the turnkey security, um, and the deep out of the box automation that these operating systems provide, they cannot be simply um, replicated by an open source community overnight. If an organization or an enterprise needs an automated highly secure fabric that just works seamlessly, then nxos is the gold standard. But like I said before, we also recognize that the future of networking like you opened up this podcast with, is inherently open, right? And our customers are operating in highly diverse multi vendor ecosystems. There is a very distinct um, Linux native customer profile out there that demands open source software control. So by expanding, so by expanding Sonic support, you know, onto our uh, flagship uh, Cisco, uh, Nexus 9K family, we are not really creating an internal collision, right? We are extending unprecedented flexibility from the silicon layer all the way up to the unified management. And the technology is very tightly integrated. But if you look at it yet loosely coupled, right? We are giving customers like an ultimate insurance policy for their capital expense expenditure, right? They can confidently deploy world class Nexus hardware today. They can run NXOs or ACI, and if their corporate strategy shifts towards open source now, or say two years from now, then they have the software agility to pivot their architecture without a physical rip and replace. So that's uh, it's like we treat our customers like true partners. So uh, there is not like that's how we make the judgment or make that call.

Speaker C: Yeah, I think that's important. Um, I really like your um, present your explanation of how you come in with an xos. I truly firmly believe that if you don't have a strong use case for open source, don't use it. Um, and I like how y' all approach that with allowing, you know, us to choose an XOS when we need to. So all the stuff across the infrastructure that, you know what, we know what we're doing, we handle it, we got it smoothed down. You can just throw an XOS at it and not worry about it. But those corner cases, those areas of the business that we still can't get to if it makes sense to bring in um, more Eyes on it, more depth into what we're working on, more architecture, more details that NXOS just can't cover across the board. Um then yeah, it makes total sense to me there and I do think that's one of those where I'm excited because I do remember the SDN days and I remember back in the past when it was like, all right, you know, all of these servers out there, they can throw whatever operating system on they want but then in networking we're kind of limited here. So it's nice to see that that's been maturing in the industry this whole time.

Speaker D: Yeah. And honestly like you know, I'm talking to a lot of variety of customers if you will. The enterprises, right. There are some um, who have a mandate uh, to go for like multi vendors ecosystems. Right. And for that reason they are asking, hey, it's more like a Sonic sticker on the switch. Hey, do you support Sonic as well? Then there are some who are trying to have some flexibility. Say they have leaf and spine in the architecture and some leaf switches. They want Sonic, they're looking to bring the cost down. You know I'm hearing a lot of um, but they're willing to make the trade offs. Right. They're very well or it's our job to educate them that it's not simply uh, uh, can I convert from uh, or migrate from NXOS to Sonic and get the same level of performance and what I was used to. No, that's not going to be right. I mean nxos is like I said, a gold standard, highly customizable, feature rich turnkey, uh, uh, operating system. Um, we are doing a lot of education and um, in helping customers make that conscious choice.

Speaker A: One of the things that enterprises are trying to do right now is build out their next generation AI data centers and there's a lot of different combinations of hardware and software that you could possibly put together. And I know that clients are often looking for help trying to figure that out because when you're making a capital expenditure of this nature, you don't want to make the wrong move. If you do, I mean bad things are going to happen. Uh, how does Cisco help these teams kind of figure out where they need to be knowing that there is a wealth of opportunity out there to pick from.

Speaker D: Yeah, I mean this is where the strategy gets incredibly exciting Dom. I mean because you know we are tailoring open source software flexibility to uh, uh, for Cisco it's three very distinct customer operational styles. Right. And think of it as three clear pillars. Right. First we have the Cisco 8K, right? Cisco 8000 and that is our powerhouse powered by silicon one, right? And that is built specifically for, so if the customers are looking for the disaggregated uh, stack, like a DIY stack, the hyperscalers, the ultra high density core transport and web scale wan backbones, you know, um, so Cisco 8K is designed for that massive scale routing tier and that supports Sonic. So they have to make a call with that use case. Second is our flagship uh, Nexus 9K uh series, right? Which is the heart of, I mean think of it as rest of the cloud, right? So which is the heart of customer enterprise fabrics and dedicated AI clusters. So what makes N9k so unique here is deep silicon diversity. So depending on the workload a customer can deploy Cisco cloud scale Silicon, Silicon One, or utilize platforms which are powered by Nvidia Spectrum X Ethernet switch silicon for really highly specialized backend AI fabrics. So this gives infrastructure teams a completely consistent um, hardware layer across a wide range of leaf spine and AI ML M topologies or architecture architectures. Right? But then we have our third pillar which is the Cisco Nexus Hyper fabric, right? And this is where we show ah, really immense humility regarding the operational realities of the enterprise. Right? Let's face it, many organizations, they want the benefits of an open standardized architecture but they completely lack the internal expertise to run and automate Rawsonic at. This is something that uh, we also talk to customers about. They go all the way into the discussion about Sonic support. But when it comes to their own internal expertise to run and automate that raw Sonic at scale, it's lacking. Right? So Hyper Fabric solves this by using Sonic underneath to blend our trusted hardware into a cloud managed vertical stack. So it completely abstracts the complexity, taking the AI fabric deployment from a massive um, multi month engineering project down to a matter of days for an IT generalist, right? So we are matching technology to really the operational DNA of the team who's holding the keyboard. Um, so these are the three core pillars around which we are um, which we offer the flexibility and choice um, to the customers.

Speaker B: I do think that integration with Spectrum X, you know, when I first started reading about this before Nvidia GTC, back in February 25th I saw a series of Cisco press releases that talked about partnership um, with Nvidia. And of course everybody likes to talk about their partnership with Nvidia. Um, and so um, I very sarcastically just kind of wrote it off at first but then I went back and actually read them and I saw okay, this architectural sharing of Spectrum X Technology. This is a level of partnership and integration that I don't think many others can speak to. Um, and it gives you um, great strength in being able to leverage open networking with Sonic and use things that fit Nvidia reference architectures very closely, uh, to be part of that infrastructure in a big way.

Speaker D: That's very interesting to hear that Scott, from you. It's a win win really. Right. Um, customers who um, are looking for that NCPRA compliance, um, and at the same time they want the power of um, um Cisco's um, hardware.

Speaker A: Ryan, does that track with some of the stuff that you're seeing out there as well?

Speaker C: It does. And a lot of what um, we hit on here, that, that resonates with the messaging I see is you know, we have a ton of trust issues out there right now with AI, um, and reasonably so we don't know where any of this is going in the next five, 10 years. We have guesses but no one has a crystal ball on it. And I think the customers that are in the point to need to make a decision right now are the hardest ones um, to do because there isn't clear solutions. So when you can bring stuff in like a Nexus platform that has that type of flexibility, we might design around heavy inference right now, but in a year we might recognize that we need to shift over to more applied AI or more of research, um type work, um training. And so being able to adjust and work with your hardware and your hardware vendor to tailor that out makes a lot of sense right now.

Speaker D: Yeah. I want to quickly ask, are you also seeing this issue of like you know, lack of skills or um, expertise in the enterprises like internal teams to be able to handle that open source, uh, running open source, uh, network operating system at scale.

Speaker B: Well I would love to tackle that one. Um, I'll preface it with great power comes great responsibility.

Speaker A: Right.

Speaker B: So uh, not to overquote my Marvel Cinematic Universe friends, but um, yeah, there is a skill issue. I want to call it a deficit, maybe that's a little too harsh. But um, you know, if you want the benefit of going with open networking, you have a couple options, right? Um, you either need to skill up and you need to understand what sonic package management looks like, um, or you need a strong partner that can do that for you in a way that's very similar to the highly integrated operating system plus networking platform. I um, don't think that's going to change anytime soon. On the operator side, you know, of course you're aware of pockets that are very sonic literate. You know, they come from the hyperscalers and some of the NEO cloud providers. But as this goes into the enterprise and again I am seeing much more enterprise interest. You know, this is going to be a skill and talent issue that we're going to need to address over time.

Speaker C: Yeah. And a lot from my side, um, I will agree with that. Um, everybody loves a conversation. Everybody wants to get into Sonic, everybody wants to play with it. It sounds awesome. Um, but then, yes, as your teams come up, they absolutely need to have a skill set in Linux. Um, they need to be able to restart services, they need to be able to watch or you have to provide a strong orchestration layer for them. So that, that part is, um, they learn how to use orchestration later instead of learning how to use Sonic. And those have kind of always been the two approaches that I've seen coming into this. Um, and it kind of just depends on what team's willing to go what way.

Speaker B: I guess to state it, um, maybe more positively, you know, as we see in public forums, more enterprise examples of Sonic deployment. That's the real evidence that this is getting traction. Right. Um, I could drop some names here. I won't for just the sake of caution. But we've seen more and more public presentations at places like nanog, like the network automation forum of customers that are talking about their Sonic deployment experience. So it's, again, it's not a hobby, uh, it's not just a science project anymore. It's really hitting enterprise networks.

Speaker D: No, absolutely. And what does that deployment experience, if you have to sum it up in nutshell, what is their experience like?

Speaker B: It always involves a strong partner and it's either somebody who is, um, a contributor to the Sonic ecosystem that knows how to do custom development if new features want to be deployed, um, you know, and just help with regular maintenance for enterprises that don't have those skills on board. Um, now I don't have visibility into, you know, what happens on day two and day 200, but if I'm, if I'm smart, if I'm an enterprise network operator and I make this decision, I don't want to be wholly dependent on a system integrator or other third party forever. I want part of that engagement with that partner to be training me and teaching me. These are the skills that you're going to need in your netops netengine network architecture teams.

Speaker C: Yeah. And one of the areas that I'm seeing Sonic, um, start to creep in, you can call it that, creep into teams or logically get added to the system is with dropship products. So full solutions that require built in network infrastructure, um, or built in network communications, dropping in a rack, plug it into the access switch and move on. Um, most people that I talk to that are really interested in Sonic fall into that space where they're solidifying their product offering and need tighter control with their whole stack.

Speaker A: Speaking of the whole stack, one of the biggest challenges that I think a lot of people are facing when it comes to embracing open software has been the question of support. We know that the best place to find support in the middle of an outage at 3am is often not on the Internet. And uh, there are a lot of companies that feel good that they can work with another company to maybe get the kind of help they need when they need it. How is Cisco changing the conversation around Sonic when it comes to getting a full featured support? When it comes to the complex interactions that happen in the enterprise and specifically in the data center?

Speaker D: Yeah, I mean um, uh, Scott brushed on it as well Tom. I mean you have articulated the exact fear I think that keeps the CIOs awake at night. Um, the distinction between um, successful deployment and a ah, failed one isn't just about the openness of the code, it's about the operational maturity of the deployment. Like you said, if a critical network issue compromises um, the customer's business infrastructure at 3am a community forum isn't going to rewrite a patch for them. Right? When customers evaluate Sonic um, they're essentially choosing between two paths, right? One is the risky DIY community path or a validated supported path. There is no other way. So buying Sonic from Cisco is really a strategic ROI decision honestly to gain the immense benefits of open source innovation um, while completely offloading the operational risk to a uh, proven enterprise buffer. So every single deployment of Cisco Sonic is fully backed 247 by Cisco's TAC, um giving enterprises that single trusted partners protocol. Right? And also we provide full integration with um, Cisco Dashboard for um, automated bring up, uh, unified life cycle management and high fidelity uh, health telemetry. Um and we are allowing enterprises to adopt an open source operating model on premium hardware they already trust seamlessly merging that complete freedom of open networking with if you will bulletproof guardrails of Cisco's global ecosystem. So um, uh, I have talked to the count teams and to the customers who are like yes we want Sonic um on say Nexus, but then uh, we don't want to tackle or hassle uh ourselves with the day two operations um, so that's what we provide Cisco support, um, and uh, that's what they sign up for as well.

Speaker A: Scott Ryan, what are your feelings when it comes to companies uh, that want to offer support for open networking? What are the things that you typically look for in those arenas?

Speaker B: Well, I want competence. Right. They understand the product combination hardware and software that they're dealing with. And there is a nuance here not to get too far into the weeds, but because you're supporting Sonic on multiple hardware platforms with different forwarding ASICs, there's this little piece of magic with Sonic called the switch abstraction interface or the sigh that has to talk to the Broadcom silicon, the Silicon One silicon or whatever else is lying in the actual forwarding plane. And the fact that you've intentionally supported multiple forwarding platforms I think shows a commitment to it and um, a knowledge of how that works that hopefully you're going to fully train and push into your ATTCK organization so they know how to troubleshoot. Yeah, when this happens I've got to look at this underlying the lowest Sonic layer. Um, I'm sure having been a TCK engineer, that's close to my heart and I uh, feel for all your TAC engineers who I'm sure are excited to learn about the next new combination of stuff going into their customers networks.

Speaker D: Yeah, no, bang on Scott. I mean because we build directly into the mainline repository and strictly adhere to uh, standard community sigh baselines, that's how we ensure that highly predictable standardized software environment. So if an issue hits a heterogeneous environment, Cisco Tag completely owns, isolating and resolving the issue on our distribution and our hardware and because the code is open and standardized, really identifying whether a friction point is at the protocol level or the physical level is actually cleaner than it ever was in um, ah, old proprietary multi vendor silos.

Speaker A: I want to close out here by maybe asking about the future because that's something that a lot of people are starting to look at now. There's a timeline for implementation of a lot of this technology. Uh, AI has really forced a lot of companies to ask hard questions about the way that they want their networks to look for the people out there that are maybe starting that, that journey where they're maybe planning out the next three to five years of the AI era. What are the things that they should be focused on when it comes to figuring out how their network should look and how open things should be?

Speaker D: Uh, I mean I can take this and I would love to hear Scott and Ryan's thoughts as well. See um, Tom, if I'm sitting down with an IT executive, right, who's mapping out say a uh, three to five year infrastructure roadmap, my strongest advice is very simple, right? Match your network operating system to your engineering team's core skill set and operational model, not industry buzzwords. Do not jump to an open operating system just because it sounds modern if you lack the internal expertise to run it at scale. Okay. So we still do and will continue to sell NXOs and ACI aggressively because for many organizations a highly integrated single vendor stack is the right business choice. Um, and we are going to, we do like on Cisco 8K and I mentioned about Nexus Hyper fabric and we will soon have support of Sonic on our Nexus family as well. But the ultimate guiding principle is to choose hardware platforms that do not lock you into a single software destiny. So by standardizing you standardize on premium Cisco hardware network architects can run different operating systems across a unified architecture silicon which is really optimized for traditional and AI traffic. So my advice is like that gives customers the ultimate flexibility to optimize for cost for performance and operational model, um, while completely avoiding uh, fabric fragmentation if you will. Right. So whether they want to choose ACI or nxos or open source Cisco, Sonic, we are delivering a comprehensive um, zero compromise portfolio that lets them scale safely.

Speaker B: Yeah, I would take everything that you said and I would add innovation onto it. Um, and in the very specific following context the need for GPU capacity is not going to end anytime soon. Um, and we see the Neo clouds emerging to support this and it's almost unfair to just talk about NEO clouds as a single entity. They're all doing different things um, and have different multi tenancy requirements that I think are going to drive some more innovation at the networking layer. Sonic gives them an option to innovate very quickly. For those that have the capability to do stuff um, with their own code and accept the risk. Um, so that's an important ingredient I think in what's going to be happening over the next 18 to 24 months, if not longer.

Speaker C: Yeah, and to your point there Scott, you know with, with the openness of, of Sonic, you're, you have multiple groups and organizations just like Cisco contributing and individual teams as well, adding to that as, as they see stuff coming up. I like how it's progressing.

Speaker A: Well as you can tell from the conversations that we've had today, there are a lot of options when you decide how you want to build the AI network of the future and the research that you are doing now is what is going to matter most in the long run because you want to make sure you're making the right choice. But just like any choice that we make when it comes to professional or personal, you really should be relying on someone who understands your needs and understands where you want to go with it. And I think one of the big takeaways from today's podcast is Cisco is poised to give you the right answer for what you need. Not necessarily selling you a specific product or a specific system, but helping you understand where you're trying to go and what the best way to do that will be. If people want to go and learn more about what Cisco is doing around Sonic specifically, but an open networking in general, where can they go to learn that survey?

Speaker D: Um, Tom, we can share that link in the show notes, um, and users can, um, you know, just click on that and there's ton of resources, um, and you know, I think the benchmark performance numbers as well. Uh, if they want to take a look.

Speaker A: Outstanding. And before we go, uh, our other guests. Scott, if people want to connect with you and continue the conversation, where can they go to do that?

Speaker B: Please reach out on LinkedIn. Um, send me a DM. I'm happy to connect.

Speaker C: Yeah, LinkedIn's usually the best place, but I also have my own, um, portfolio and blog out at Abstractrian AI.

Speaker A: And lastly Serbia. If people want to connect with you and learn more about what you do, where can they go to do that same?

Speaker D: LinkedIn. I think that's the only social media platform I want, so LinkedIn is the way to go. Please feel free to reach out. Happy to talk.

Speaker A: Thank you all very much for listening to this episode of the Tech Field Day podcast. If you enjoyed this discussion, please do us a favor and subscribe on YouTube or in your favorite podcast application so you don't miss an episode. Consider giving us a rating and a nice review because that does help help us find new viewers. This episode was brought to you by Cisco and Tech Field Day. Ah, the home for IT experts from across the enterprise. Tech Field Day is a part of the Futurum UM group and for more upcoming events and episodes, head over to techfieldday.com podcast or check us out on Tech Strong TV. Thank you very much for listening and we'll see you all next week.

Speaker D: Sa.