Nico Vibert is Director of Technical Advertising Engineering for Isovalent at Cisco. Isovalent, acquired by Cisco in 2024, leverages eBPF, Cilium, and Tetragon know-how, which have change into de facto constructing blocks for cloud-native networking infrastructure.
To learn the way Cisco is extending cloud-native networking performance throughout enterprise networks and cloud materials, I not too long ago interviewed Vibert about the advantages of Isovalent know-how and the way it suits into the Cisco portfolio of networking, observability, and safety.
Integrating the enterprise and cloud material
Vibert had many distinctive insights about how Isovalent can be utilized throughout the Cisco portfolio, from enterprise information facilities to the cloud.
First, some background. Isovalent was created after Kubernetes and containers turned customary infrastructure within the cloud. It turned clear that container networking and cloud-native networking wanted a distinct community and safety mannequin.
Cisco had the foresight to buy Isovalent in 2024, buying the chief within the container and cloud networking house. Isovalent makes use of open-source eBPF know-how to construct Cilium, one of the vital common instruments for container networking in cloud-native environments.
As a result of Cilium is constructed into the Linux kernel, it could possibly allow engineers to leverage fine-grained safety and community properties with out impacting efficiency.
“Cilium actually took off and have become extensively adopted within the container house and Kubernetes house to the purpose the place it was utilized by all the foremost cloud suppliers, like Google, AWS, and Microsoft Azure, for Kubernetes choices,” Vibert informed me.
Vibert stated that Cisco now has a novel management place amongst networking information facilities and cloud infrastructure, together with AI infrastructure. By melding cloud-native software-based networking capabilities with networking underlay, the corporate can ship full-stack safety and observability for any sort of community, together with hybrid and multicloud networks.
“For me, it’s the perfect of each worlds,” stated Vibert. “You deliver Kubernetes networking experience scalable, high-performance fashionable workloads. After which you have got Cisco bringing many years of expertise Like Cisco Nexus One working high-performance vital purposes within the information heart.”
A unified operational expertise
The updates to Cisco Nexus One, which I’ve highlighted on this weblog, give Cisco distinctive capabilities to construct a constant operational expertise throughout on-premises, cloud, and API-driven environments.
Vibert describes it as a layered method, with Nexus One doing the heavy lifting for the networking underlay and bodily infrastructure and Isovalent doing the work on the larger layers of the software program stack—with further visibility on the operating-system stage.
“You could have Nexus, which offers the inspiration,” says Vibert. “It offers predictable efficiency, reliability, and safety on the bodily stage. Isovalent offers enterprise networking prolonged to containers, the cloud-native world. This offers constant networking and safety for Kubernetes, the place you may put repeatable code-driven operations,” continues Vibert. “We configure Nexus as infrastructure as code and it’s all coverage pushed. We’re making use of the identical mannequin in Kubernetes, and it’s pushed from code with enterprise logic to deploy networking from intent.”
What Vibert describes is the holy grail of networking know-how, utilizing a software-defined method to tie collectively conventional enterprise underlays with cloud-native materials.
Potential for AI environments
As AI workloads demand unprecedented scale and deterministic community efficiency, this method will change into helpful for scaling the immense measurement and calls for of Kubernetes clusters, that are key to AI purposes.
Kubernetes shortly turned the common platform to face up and deploy fashionable cloud purposes, and AI isn’t any completely different. Whether or not it’s LLM coaching, inference, or agentic AI, Kubernetes clusters and information will proliferate within the AI world.
“There are case research of Kubernetes environments with 1000’s of nodes, proper?” stated Vibert. “Kubernetes has change into the platform to run AI workloads. Within the conventional world, whenever you run high-performance computing, you want a really robust networking basis, so it’s the identical in Kubernetes. That’s the place we excel. Cilium is being utilized by a number of the largest AI and enormous language mannequin (LLM) suppliers on the earth, and one of many causes we’ve change into profitable is with eBPF.”
One other space the place Cisco thinks Isovalent will assist AI is by lowering the price of managing the infrastructure to attach GPUs. With a major funding in AI infrastructure and a compressed timeline for execution, it’s necessary that AI information facilities get related shortly, on the lowest doable value. Isovalent offers a mechanism to attach AI and Kubernetes pods with multi-tenant safety and excessive efficiency.
“You must have a safe multi-tenancy platform,” stated Vibert.
Vibert says that eBPF and Cilium are excellent for this job as a result of they’re constructed into the Linux working system on the kernel stage, offering programmability and efficiency advantages. Utilizing Cilium and eBPF, Isovalent can add networking features into Cilium with out compromising efficiency as a result of most of them change into kernel features.
“Cilium is letting us do networking features at kernel pace,” says Vibert. “That makes an enormous distinction for LLM firms that may’t have bottlenecks within the networking.”
Addressing cloud device sprawl
Vibert stated that Isovalent’s Cilium and Tetragon can even assist by addressing cloud device sprawl and integration complexity. He believes device sprawl is pervasive in cloud environments and contributes to prices. The rise of AI has the potential to exacerbate the issue.
“What occurred over time is that platform groups began to undertake increasingly more single-purpose instruments for plenty of completely different use instances, together with for safety, for encryption, and for observability. There can be particular person instruments deployed for service mesh and proxies.”
Isovalent has already helped many shoppers cut back device sprawl in hybrid and cloud environments, together with a significant monetary agency, in accordance with Vibert. It offers enterprises the real-time, actionable visibility they should transfer quick with confidence, turning reactive firefighting into proactive optimization.
“What organizations must do is have the ability to benefit from [their resources]. You possibly can spend weeks constructing an AI app to have it break. That’s why we’re going to be robust at Cisco round observability since you’ve bought Cilium and eBPF working at a deep stage to know community efficiency to see each packet throughout the platform.”
In abstract, Cisco Isovalent gives excessive efficiency and safe cloud-native networking that connects your workloads and infrastructure throughout Kubernetes, cloud, information facilities, and legacy infrastructure.
As Vibert has outlined, Cisco is off to a robust begin with the Cisco Nexus One extension of infrastructure materials to cloud-native Isovalent networking. The usage of eBPF, Cilium, and Tetragon will present seamless connectivity with cloud-native environments, fine-grained community visibility, high-performance safety, and a streamlined operational expertise with each networking and cloud groups.
