Hello everyone and welcome. In this video, we'll explore how Cisco AI defense is deployed on Cisco AI pods. We'll cover the platform architecture, infrastructure requirements, deployment process, operational considerations, monitoring capabilities.00:04And the best practices for running AI defense on Cisco's AI optimized infrastructure. We'll start with an overview of Cisco AI pods and explain why they provide an ideal platform for on prem AI defense instance. From there, we'll examine the architecture.00:19Review the open shift platform requirements, prepare the environment, walk through the deployment process, and finish with operational guidance, monitoring, and best practices. Cisco AI pods are validated reference architectures designed specifically for AI workloads. They combine Cisco UCS compute platforms and.00:36Nvidia GPUs acceleration.00:56High performance networking, enterprise storage and integrated management into a scalable solution. Rather than designing AI infrastructure from scratch, organizations can deploy a validated architecture that's optimized for AI training, model inference, fine tuning, and security workloads such as Cisco AI defense.00:58AI initiatives rarely stop at deploying a single language model. Organizations need infrastructure that can support model training, fine tuning, inference, and future AI growth. Cisco AI pods provide that validated foundation while Cisco AI defense extends the platform with built in security.01:17Enabling organizations to confidently deploy and protect AI applications throughout their life cycle. Many organizations host their own AI models to meet data sovereignty, regulatory or performance requirements. Deploying AI defense directly on Cisco AI pods allows security inspection to happen close to the AI models themselves.01:35Reducing latency while ensuring sensitive prompts and enterprise data remain inside the organization's environment. The Cisco Security Cloud hosts the management plane while the AI defense data plane runs inside the customer's open shift cluster on AI pod infrastructure.01:55This architecture separates management from data processing. Administrators manage AI defense centrally through Cisco Security Cloud while AI traffic remains inside the customer's environment. Once the infrastructure is ready, deploying AI defense becomes a straightforward Kubernetes operation.02:11Instead of installing a monolithic application, OpenShift automatically deploys operators, microservices, persistent storage, and supporting components, making the platform scalable, resilient, and easier to maintain. Cisco validates AI defense on Red Hat openshift.02:30A production deployment requires at least three control plane nodes to ensure high availability. Worker nodes host AI defense workloads and GPU resources. Open shift operators automate infrastructure management while Kubernetes scheduling distributes workloads across the cluster. GPU operators ensure AI workloads can efficiently access GPUs while Kubernetes schedule.02:47Tools workloads across the cluster for high availability.03:07Before deployment, the AI pod environment must satisfy infrastructure prerequisites, DNS resolution, outbound HTTPS connectivity, and video GPU drivers, minimal open shift version and operators persistent storage and certificate management must all be validated.03:12Completing these prerequisite checks before deployment significantly reduces installation time and helps avoid configuration issues later in production. Deployment begins within Cisco AI defense SaaS instance, where administrators generate an AI pod deployment script. Kubernetes manifests are then applied to the OpenShift cluster. Operators deploy.03:29The required microservices, initialized storage, configure networking, and register the deployment with Cisco AI defense. Administrators verify pod readiness and cluster health before enabling production traffic.03:49After deployment, AI defense consists of multiple specialized services. Gateway services, process incoming requests, policy engines evaluate security rules, detection engines, identify prompt attacks and model vulnerabilities. Together, these services provide a layered approach to protecting AI applications while continuously collecting operational and security telemet.04:03For monitoring and analytics.04:24Daily operations involve monitoring service health, updating security policies, managing software upgrades, and validating cluster capacity. Because AI defense on AI pods is fully Kubernetes native. Administrators can rely on familiar open shift operational procedures while managing AI specific security policies through Cisco AI defense.04:27Observability combines Kubernetes native monitoring with AI specific security telemetry. Cisco AI defense complements these capabilities with AI focused security events, policy violations, blocked prompts, and vulnerability findings. AI defense is designed for enterprise availability. Kubernetes repli.04:49Because ensure service continuity, anti affinity distributes workloads across nodes and additional worker nodes can be added to increased capacity.05:09GPU scheduling enables AI workloads to efficiently utilize available accelerators. Many enterprise customers integrate Splunk Enterprise with their open shift environments. This provides a centralized view of AI defense services, Kubernetes resources, cluster events, infrastructure performance, and security alerts from.05:17From a single operational dashboard.05:37While OpenShift already provides native monitoring through Prometheus and Grafana, Splunk Enterprise extends observability by consolidating infrastructure logs, Kubernetes events, AI defense telemetry, and security alerts into a single dashboard. This enables operations teams to quickly identify issues, investigate incidents, and monitor the overall health.05:40Off of the AI pod environment.06:00A successful AI defense deployment begins with careful planning and selecting the right AI pod configuration. Customers should evaluate the size of their AI models, expected application traffic, concurrency, GPU utilization, and future growth before choosing an AI pod configuration.06:04Cisco offers multiple validated AI pod architectures designed for different AI workloads from inference and rag applications to large scale model training and fine tuning. Once deployed, Cisco recommends using dedicated worker nodes for AI workloads, persistent storage for stateful services.06:20Continuous monitoring of GPU and cluster health, regular software updates, and following Cisco validated designs to achieve optimal performance, scalability, and long term supportability. Cisco AI pods provide a validated enterprise platform for deploying and securing modern AI applications by combining.06:38Learning high performance AI infrastructure, Kubernetes native deployment and Cisco AI defense. Organizations can confidently adopt AI while maintaining security, operational efficiency, and enterprise scalability. Thank you for your attention.06:58
