Why Broadcom Is Betting on a Private Cloud Comeback  🔗 **Source:** [Why Broadcom Is Betting on a Private Cloud Comeback](https://share.snipd.com/episode/24f5ff56-98de-446d-ab9a-997bff3f4ff0) ## Highlights - **Private Cloud Is Making A Comeback** - Private cloud is re-emerging because enterprises want data sovereignty and AI workloads kept on-prem. - Jad explains enterprises are bringing cloud-native stacks back on-prem while keeping community open source participation. Transcript: B. Cameron Gain Who can say that? Yeah. Jad So, I mean, this is a, we could probably fill up a day. High level. A day on this content. But, I mean, in short, look, we've been around the block. We've been around this industry for quite some time. We've seen the shift from this pure infrastructure play, the massive private data centers, the enterprises wanting to maintain and control all of that to this shift to the hyperscalers And then eventually the cloud native services and tying their apps to that. And we're somewhere in between those two over the last decade or so. But more recently, we've seen not only enterprises shift to the private cloud. But also the willingness to incorporate open source and CNCF. Now, that in itself isn't brand new, but the willingness for an enterprise to make that as part of the native stack in the enterprise services and then bringing that as a second part of That to the private cloud, we personally, or I personally, have been seeing more and more and more of that. And back in my enterprise days, OpenStack was where you practiced, right? Yep. Today, it is core enterprise infrastructure stack and services stack. So there's a lot of things that are happening there. I kind of mashed a bunch of it together. It's not new, but the momentum is building somewhat. And the momentum is obvious, because you are now at a point where you've got these services, you've got that infrastructure, you're part of the community, but also a big part of the consumption. And now you have AI and you have new concerns over data sovereignty, new concerns over infrastructure, and who owns that infrastructure, who has access to that infrastructure. And what we're seeing is that stack that has now become cloud native and the enterprise fully adopted it is now coming back on-prem. ([Time 0:02:09](https://share.snipd.com/snip/286a01a0-5f0f-4c73-bda1-903722d70e13)) - **Open Source Is Enterprise Grade** - Open source quality now matches enterprise needs so customers demand optionality over proprietary lock-in. - Miles says customers ask about standards like OpenTelemetry, so VMware is incorporating and contributing to CNCF projects. Transcript: Miles Some of those points. So basically, like Jad was saying, there's sort of a stack that people are using of tooling. And traditionally, people were using tooling that was first-party vendor proprietary software, because that's where the quality was. That's where the features that they wanted were. And open source is of such high quality nowadays that you don't necessarily need something completely proprietary to do that and that's why we're trying to you know bring on more and More open source projects because if we go and talk to customers our customers are talking to us about you know what is your story for open telemetry or you know like so these uh these protocols These projects are the standard now it doesn't matter that you built something proprietary and internal people want optionality they don't want vendor lock-in and all of that comes Back to you know the quality of open source is so high that it really doesn't make a ton of sense to go and reinvent the wheel maybe the best thing that you can offer is support packaging That, that kind of stuff. ([Time 0:04:30](https://share.snipd.com/snip/775c4cd8-5564-47da-bad4-b805a38b9c80)) - **Donating Velero And Contour Built Community Trust** - VMware has donated projects like Velero and Contour to CNCF to increase community collaboration. - Miles and Jad cite multiple donations and contributions to SIGs as evidence of VMware's pro-community stance. Transcript: Miles And it's not just Valero. We have Domated Contour as well and there's Harbor and there's a whole bunch of first party open source that we've helped bring to bear on that too. Jad And arguably I think Miles touched on the trust of the open source community, but that's happened largely because of the open source community's alignment with the enterprise. I mean, when we were first using open source, again, it was in labs. B. Cameron Gain It was like a beta, a toy project. Jad Even Kubernetes was like that, where customers didn't take Kubernetes to production until several years after the wide adoption. It took them that long to figure it out. It is hard. And they're still figuring it out. It is hard. But so the thing is, when you have a company like VMware that has contributed and built into these various projects, then donate those to the community. That is one example of how you take a company, again, like VMware, so enterprise aligned, taking that DNA and putting it out in the community. This is what we're seeing across the board from several of your traditional enterprise-type services pushing out these projects in the communities. B. Cameron Gain How does that manifest itself when you do that? I didn't quite get that. Are you, is that the form of more commits, pull requests, or for the projects? Or how is that? In some cases, yeah. Miles You know, some cases will contribute fixes to upstream and that kind of thing. You know, we contribute to a number of the SIGs in the Kubernetes community to try and help drive those projects. ([Time 0:05:59](https://share.snipd.com/snip/1d8cd030-936c-4c82-82e8-d3391a27038d)) - **Heptio Acquisition Shaped VMware's Kubernetes Path** - VMware's 2018 Heptio acquisition marked a pivot toward Kubernetes expertise inside the company. - Jad recounts how internal container projects and Heptio, Pivotal ties coalesced VMware's Kubernetes strategy. Transcript: Jad So that happened in 2018, the Heptio acquisition. And that was a very important pivot point, I think, for VMware. But the strategy behind it actually started before that. We had internal projects where we're running containers on vSphere and our hypervisor, because, know it kind of made sense to do that you've got this gold standard of a virtualization Platform that the vast majority of the enterprises in the world were running and you're like well let's just extend that to containers and so that strategy started way back then and Then the acquisition of Hept, our involvement with Pivotal and Cloud Foundry and all these different pieces kind of coming together and started bringing different communities together, But impacting the platforms as we went along. ([Time 0:08:56](https://share.snipd.com/snip/b3aff44b-ad59-4d7c-81e5-7a368e677af4)) - **Kubernetes Is Now The Platform Control Plane** - Kubernetes is not a bolt-on; VMware rebuilt platform APIs around Kubernetes control plane and kubectl/kube API. - Miles notes provisioning (storage, networking, compute) now happens via CRDs and Kubernetes declarative APIs. Transcript: Miles Know, the Kubernetes stuff is not just a bolt-on. Like, you know, it could be perceived that it's just another workload. We just threw Kubernetes in there and that's what you get with the platform. We fundamentally changed how you provision and use the APIs to create those workloads. So we actually, we have this modern API, like unified API surface now that we use internally in our product that is KubeCTL and KubeAPI driven. So we're provisioning Kubernetes clusters using Kubernetes CRDs. That is our control plane now as well. So all of our new integration services projects, they are all operated through KubeAPI as well. So it's not just like we have Kubernetes as an add-on. ([Time 0:12:27](https://share.snipd.com/snip/c1873d04-8ec3-42bc-9992-7aea001d687f)) - **Apps Should Transcend VMs And Containers** - Distinction between VMs and containers is fading; customers want a single platform for all application types. - Jad argues the private cloud should offer a unified pipeline like public clouds where apps span VM and container runtimes transparently. Transcript: Jad The one platform, right? So, I mean, traditionally, you're going to have your PaaS or AppStack, and you're going to have your InfraStack, right? And different players play different roles in those, and a lot of intersections and so on. But more and more often, we're seeing, from a runtime perspective, you just, I mean, Kubernetes is infrastructure. Containers run on infrastructure. VMs run on infrastructure. And infrastructure is kind of the foundation for all of it. So we see this relationship brewing, but also coming to fruition where customers are saying, look, why do we have to pivot between these two very disparate platforms to run apps, what We call modern apps, a term I absolutely hate, and our traditional VMs, another term I absolutely hate. It's all apps, right? And why are we pivoting across two platforms? We're at a position now where we need a platform. So when I go to AWS or Google or Azure or any of those, I'm not creating lines in the sand and how I consume those clouds. I've got my EC2 and I've got my EKS, right? The private cloud should be exactly like that. And whether it is a whole set of VMs or a whole set of containers or Kubernetes clusters or whatever it is, it should be done in one pipeline, one workflow, one set of flow, because your Applications actually span all of that. ([Time 0:13:30](https://share.snipd.com/snip/d8368e45-7459-45d7-8471-997b8c65dcb0)) - **Use VCF As The Unified On-Prem Cloud Platform** - Treat VMware Cloud Foundation as a unified platform with vSphere, networking, storage and VKS layered together. - Jad explains VCF bundles core hypervisor and virtualization stacks with VKS providing an upstream-conformant Kubernetes runtime. Transcript: Jad Yeah, so we have a couple of different aspects. And so the first one is obviously bringing the whole cloud management and the whole hypervisor and hypervisor management stacks together, our network virtualization, our storage Virtualization, all just becoming core platform. And on top of that is where some of this magic happens, right? We have the vSphere Kubernetes service. It is an upstream conformant Kubernetes runtime, right? And it's also where we are shifting a lot of the management from a platform perspective on top of. It's something called the supervisor, but that's just what we call it. But it's really just a Kubernetes control plane and that declarative API and so on. VKS. B. Cameron Gain It's a shift of engineers towards that Kubernetes control plane. Yeah, for our product. That's what I was alluding to earlier. Miles Our product is being built around those APIs as well. Yeah. Jad So whether you're provisioning storage or networking or compute or a container, I mean, a lot, that Kubernetes style API, the declarative API is going to be your one-stop shop. Now, just to put an analogy there, right, again, Amazon, you've got EC2, you've got EKS, Azure, I mean, you've got your, what is it, the VM service and the AKS. VMware, you've got your VM service, which now runs on top of a, or beneath, or on top of a Kubernetes API. And then you've got your VKS, which is your Kubernetes, and then container and everything else on top of that. So yeah, it just comes together. ([Time 0:17:33](https://share.snipd.com/snip/a68cf974-42bc-4278-aa80-f42ff327ad9c)) - **Broadcom Chose Private Cloud Focus Over SaaS** - Broadcom intentionally deprioritized SaaS to focus engineering on delivering a private-cloud-first offering. - Jad says the company pulled back some SaaS to concentrate on private cloud operating model customers request. Transcript: Jad Pretty much, yeah. And that extends to SaaS also. Like, Broadcom made an early, post-acquisition of VMware, made an early decision that, you know, SaaS is not our business model because we're seeing what customers, or we're responding To what customers are asking for. Like, listen, we want the private cloud, but we want to operate it like a public cloud. We want all the same constructs, but we also need this Kubernetes dial tone. If we were to continue to do this whole split brain SaaS and on-prem, it wasn't giving us the focus that we needed as an engineering organization to deliver the goods. And our leadership made that decision early on. And I say this because it's come up a lot. How come you pulled back all your SaaS services? And I said, but that's the strategy. And that's what our customers are asking for. And we're responding to that. It helps us focus. It helps us be the dominant player in this private cloud. There's only a handful. And so, yeah, I mean, that's what my point is that SaaS extends the same concept here of coming back on-prem. ([Time 0:20:51](https://share.snipd.com/snip/382a33dd-c476-4f43-a9cb-675bf59c853f)) - **Drive Platform Delivery With GitOps** - Use GitOps and the Kubernetes API surface to automate cluster and app delivery instead of manual UI workflows. - Miles recommends driving operations through CRDs so tools like Argo CD or Flux can be used for platform automation. Transcript: Miles Again, you drive all this via the Kubernetes API. We're not shy about saying that we think the way that you should run these modern workloads is through a GitOps practice. You should be driving it through CRDs. That's why we've written all of our API surface is through the Kubernetes API to allow you to do that, right? So yes, we realize the UI clicky-clicky thing is important for a lot of customers, but once you get to a level of maturity, you probably want to automate that and GitOps, you know, we see Is the way to do that. And the best way to provide for that is to provide that Kubernetes API surface so you can use standard tools like Argo CD or Flux or whatever. ([Time 0:22:32](https://share.snipd.com/snip/12d68be1-322a-426f-abda-c251255959b1))