The Cisco Data Center Architecture in 10 minutes [2025]

[00:00:00] Speaker 1: - In just a few years, data centers have changed dramatically. They're no longer just on-prem. They are hybrid, they are at the edge, and they are powering AI. Modern applications need speed, scale, and smarts. And businesses need consistency, security, and control. And today, on-prem infrastructure has to feel simple and seamless, just like in the cloud. The good news is that Cisco delivers. Let's take a fast lap through our architecture and see how it can help you innovate, scale, and stay secure. This is the Cisco data center in 10 minutes. Think of your data center network as a superhighway. And what about modern apps? Well, they are the sports cars. Classic layer two networks with 10 gig interfaces just can't keep up. And this is where Cisco Nexus 9000 switches and Nexus Dashboard are game changers. With a diverse portfolio of Cisco Nexus 9000 switches, you can achieve speeds of up to 800 gig per interface. With Rocky version two support, providing lossless ethernet. And the flexibility and performance needed for AI and traditional workloads is just there. Then, with VXLAN running between spine and leaf switches, and VPCs for your server connections, you get unlimited bandwidth, scalability, in an active-active architecture, seamless VM mobility, even across sites, and consistent traffic segmentation. But let's face it, manually configuring VXLAN on every switch is complex and time-consuming. With Nexus Dashboard, which is included with every Nexus license switch, configuring your network is as simple as selecting a fabric type, like VXLAN or AI. It automatically discovers and configures your switches. And then, for ongoing management, Nexus Dashboard lets you centrally manage two, ten, or thousands of Nexus switches across multiple sites, whether you're running NXOS, ACI, or even both. This not only allows you to extend VXLAN between them, but it also provides a single and consistent data plane with a Configure One's Deploy Anywhere model. Plus, with real-time telemetry, fault detection analytics, and AI-powered tools, Nexus Dashboard dramatically simplifies and consolidates operations for Nexus switches and other third-party and Cisco devices. But what if you prefer to have a Cisco-managed solution for your data center network instead? And that's where Nexus Hyperfabric steps in. Just model your network in the Hyperfabric portal. And then, when your switches arrive, Hyperfabric will automatically configure them using AI best practices. That means no installs, no maintenance, just a hands-off, high-performance network solution. Networking is your highway. Now, let's talk about the engine, compute. And there are three key trends driving demand. The first one is edge computing, which brings processing closer to where data is being created, like in stores or sensors. The second one is AI and machine learning, which need massive power to train models and crunch data at high speed. And the third one is traditional computing, which is usually used for core applications like databases and virtualization. But here's the kicker. Modern applications and AI workloads, whether in the data center, at the edge, or in the cloud, need more than just CPUs. Enter GPUs. GPUs are chips installed on your servers that are built for parallel processing, making them perfect for AI, machine learning, and big data analytics. Sure, you could run AI on CPUs alone, but it would be like racing a bicycle against a sports car. The good news is that Cisco brings UCS and Intersight, the perfect hardware plus software combo for AI, edge, and traditional applications. With Cisco UCS, you can abstract hardware configurations into profiles and templates, or assign the right operating system, just through blueprints, even before booting up a server. The Cisco UCS server lineup is built to win, no matter the race, no matter the workload. So let's break it down. First, UCS C-Series, our rack servers. These are the muscle cars of compute. Think of reliable, powerful, and capable of running both traditional workloads and GPUs for AI. The second one, UCS X-Series, our modular servers, the future-ready hybrid racers of compute. These are designed for space and energy efficiency, so that they can scale easily as your needs evolve. And here's where it gets really cool. With the X580P GPU module, you can now share GPUs for AI workloads, even across multiple servers. Number three is purpose-built AI server line, such as the C885 and the C845. These modular systems follow NVIDIA's HGX and MGX architectures, and are packed with GPUs to tackle the speed and intensity of AI workloads. And lastly, number four, the Cisco Unified Edge. This is the Swiss army knife for all your edge environments, combining UCS, catalyst, networking, and SD1 with GPUs and security into a compact solution. This is perfect for branches and remote sites. And powering it all, Cisco Intersight. Your cloud-based or on-prem management platform that provides you with centralized management for all types of servers, AI-powered insights to quickly identify issues and get answers, and provides a configured-once deploy-anywhere model that simplifies your operations. Did I mention that it integrates also seamlessly with Nexus Dashboard? But even the best engine needs fuel, and that's where data storage comes in. AI workloads creating consume massive amounts of data for training and inference, and traditional applications require fast and reliable access to it. So whether you go with storage arrays or hyper-converged infrastructure, which brings storage compute and networking into a simple platform, you need scalable, high-performance data access. And Cisco doesn't build storage, but we work with the best. Together, we deliver validated, high-performance solutions, whether it's on a dedicated storage array or software-defined storage on UCS. So how does storage and servers connect, you may be asking? Well, we've got a few options. First, fiber channel, with MDS 9000, which provide up to 128 gigs, and also managed by Nexus Dashboard. And two, IP-based storage, which runs on top of Nexus 9000. This is usually common for AI, delivering up to 800 gigs and lossless Ethernet. And with conversion stacks like FlexPod and FlashTech, it's all plug-and-play, faster to deploy, easier to manage, and ready for anything. All right, right. We talked about infrastructure, but where do your workloads actually leave? Well, there are three main hosting options, whether on-prem or in the cloud. One is Bare Metal, which is dedicated for performance. Two, Virtualize, which is efficient and flexible. Or three, containers, which are portable across clouds and data centers, making them scalable and ideal for AI. But managing containers at scale isn't that easy. And that's where Kubernetes comes in, to automate the deployment, the scaling, and operations. However, running Bare Metal Kubernetes can get complicated. And it's just like imagine building a racetrack every time you want to race. You want to avoid that. But here's where Cisco and Red Hat OpenShift step in. OpenShift simplifies Kubernetes deployment, making it secure, scalable, and optimized for AI. And with other Cisco technologies, such as Isovalent, that includes Cilium, Hubble, and Tetragon, we can deliver granular kernel-level visibility and enforcement for Kubernetes and cloud-native environments. Imagine now all this, together paired with UCS and Nexus, we can deliver a best-in-class conversion solution. Speed is great, but without security, it's like driving a car with no brakes. Today's threats move fast, especially in AI pipelines. And that's why security needs to move even faster. This is where Cisco Security Cloud Control and the Cloud Protection Suite come in. With a single cloud-native platform where you can centrally manage all Cisco security tools, including Hypershield, Multi-Cloud Defense, and many more. Hypershield puts intelligence security where it matters. Right at the network, cloud, or workload. And built on top of Isovalent's Tetragon, powered by eBPF and infused with AI, it continuously adapts to application behavior, enforcing micro-segmentation and protecting against exploits. And when deployed on Cisco's N9300 smart switches with DPUs, it offloads tasks like encryption, segmentation, and telemetry from CPUs, boosting performance and security at scale. And then there's Cisco AI Defense, which prevents data leakage and protects against emerging threats for both AI and LLM models. The result is a full-stack AI-ready security architecture from on-prem to cloud across traditional AI workloads, centrally managed and tightly integrated with dual-war infrastructure. Want everything we just covered pre-integrated and ready to go? Cisco AI pods are the plug-and-play solution for drivers who want full control. They combine Cisco UCS, Nexus 9000, the storage from partners such as NetApp, Pure, or Vast Data, and NVIDIA AI tools into a pre-validated package that works seamlessly with inter-site for compute management, and either Nexus Dashboard and Nexus Switches for traditional hands-on network operations, or Nexus Hyperfabric for a hands-off, fully managed connectivity solution. It's like having an all-inclusive race car that lets you focus on winning while the pit crew handles everything else. Plus, built-in NVIDIA licensing lets you launch AI with just a single click, making the creation of your secure AI factory a cloud-like experience. From networking to compute, from storage to security, and from do-it-yourself to fully managed, Cisco's data center architecture delivers speed, simplicity, and security, whether you're running AI, scaling your applications, or simplifying operations. We want to put you in the driver's seat to win. And that's the Cisco data center architecture in 10 minutes.