Insights from the Clastix team

Who controls the switch tenant traffic passes through decides your CPU bill. A look at network isolation, overlay costs, and moving the work onto DPUs.

Kamaji’s creator responds to two recent vCluster posts comparing Kamaji and vCluster. He agrees with three of their criticisms and corrects two factual errors: Kamaji’s default datastore is etcd, not SQLite, and each tenant runs on its own worker nodes with its own compute and network, so tenants never share nodes. Each correction links to the Kamaji documentation and to vCluster’s own posts.

From metal to kMetal, a unified platform built for AI workloads, supporting Kubernetes to GPU-accelerated computing, with zero-trust, cloud-like scaling, and real multi-tenant isolation. CLASTIX CEO Adriano Pezzuto explains why the next phase of AI will be decided at the infrastructure layer, and how the €2.9M seed investment from CDP, Mistral, and Vertis will be used.

For most of Kubernetes' life, a cluster meant machines. Three of them, at minimum, doing nothing but running kube-apiserver, kube-scheduler, kube-controller-manager, and etcd. Dedicated, tainted, and mostly idle: a control plane for a fifty-node cluster is not meaningfully busier than a control plane for a five-node one, yet both demand the same quorum of three machines, the same three etcd members, the same patching cadence, the same certificate rotations, the same backup and restore runbooks.

kMetal 1.0 from CLASTIX takes production Kubernetes down to the bare metal – no separate hypervisor layer, no VMware tax, no infrastructure stack bolted on beneath your clusters. This deep dive walks through the capabilities behind the platform: hosted control planes that run every tenant's control plane as pods rather than on dedicated nodes, VM-grade tenant isolation on shared hardware, self-service networking with per-tenant VPCs, secure GPU pooling across tenants, and fleet management for hundreds of clusters – all defined and version-controlled through one Kubernetes-native API. The result is multi-tenant, bare-metal Kubernetes operated as code: one platform, one API, built for enterprises running Kubernetes at scale.

kMetal 1.0 runs production Kubernetes directly on bare metal with no separate hypervisor layer – dropping the costly, overkill virtualization stack from your Kubernetes infrastructure. It brings hosted control planes, embedded tenant isolation, and fleet management together in one platform, plus secure GPU pooling to share scarce GPUs across tenants at high utilization. Efficient, multi-tenant, and fully as-code, for enterprises running Kubernetes at scale.

When a CVE affects your Kubernetes control plane, the clock starts immediately – and where you are in the patch queue depends on who maintains the software you're running. This post covers what proactive CVE support actually looks like from the maintainer's side, why commercial subscribers to Kamaji get access to stable, patched branches before public disclosure, and why that structural advantage matters more as the EU Cyber Resilience Act raises the compliance bar for infrastructure providers.

Years of open source work on Kubernetes multi-tenancy were always going to lead here – down to the metal. In this article, CLASTIX CTO & Co-Founder Dario Tranchitella walks through what kMetal is, why it exists, and what is inside it: eight components that together take a powered-on bare server to a fully governed, isolated, multi-tenant cluster fleet – with no proprietary hypervisor layer, no networking appliances, and no control plane infrastructure that grows with the fleet. Whether you are a platform engineer evaluating the architecture or a budget holder asking what this means for TCO, the answer is in the details.

In this article, we explore how Kamaji, combined with Cluster API and kMetal, enables a powerful alternative to delivering a true managed Kubernetes experience across cloud and edge, with full control, higher resilience, and without vendor lock-in.

Part 2 of our kMetal introduction. A deep dive into how kMetal works: hosted control planes, kernel-native KVM isolation, software-defined networking, fleet reconciliation, cluster autoscaling, and multi-site HA.

How we went from building Kamaji, the open-source hosted control plane engine, to kMetal, a full Kubernetes platform for bare metal with embedded isolation and fleet management. Part 1.

In one of the most insightful technical posts of the year, Landon Clipp published a brilliant deep-dive into what it actually takes to build a GPU-based Containers-as-a-Service (CaaS) platform on modern NVIDIA HGX systems.

A deep dive into Hosted Control Planes: what they are, why they matter and how to run them in practice.

The NeoCloud Opportunity: Transitioning from Hosting to Cloud-Native Platforms

In a world increasingly defined by cloud-native infrastructure and zero-trust principles, immutability is emerging as a foundational pillar for secure, predictable, and manageable Kubernetes deployments. This is especially relevant in bare metal and edge use cases, where traditional rolling upgrade strategies often fall short.

You keep wearing the same hat and wondering why everything looks the same: maybe it's time for a new fit.

Shaik Zoheb is a Kamaji project ambassador based in the United Arab Emirates. He focuses on promoting the Hosted Control Plane architecture, particularly on networking integration and optimization using Konnectivity.Here's the chat with Dario, Kamaji maintainer and creator.

Managing multiple Kubernetes clusters can be resource-intensive and complex, especially for service providers, hosting companies, or large enterprises. An effective strategy to streamline operations and optimize resources involves integrating Kamaji by CLASTIX with NetApp's Trident.

Artificial Intelligence (AI) is transforming industries at an unprecedented pace, with Generative AI driving much of the change over the past two years. For both Public and Private Cloud Operators, mastering the design of scalable, efficient, and secure AI infrastructures is crucial to remaining competitive. This piece provides a roadmap for navigating the technical challenges and seizing the opportunities in building cloud infrastructures tailored for the next generation of AI-driven technologies.

The Kamaji v0.6.0 release besides welcoming new contributors introduces exciting features and enhancements, setting a groundbreaking level to the Hosted Control Plane architecture.

Managed Service Providers (MSPs) and Cloud Service Providers (CSPs) play a critical role in the adoption of containers and Kubernetes, helping enterprises transform their legacy stacks by providing access to Cloud Native Solutions and bridging gaps between cutting-edge technology and the enterprise. CLASTIX introduces a new program for MSPs/CSPs, helping them to build leading cloud-native solutions for their customers.

Clastix announces a new Serverless GPU solution developed in partnership with Seeweb, Italy's leading provider of cloud infrastructures for Artificial Intelligence.

As the field of Artificial Intelligence (AI) and Machine Learning (ML) Operations matures, it is necessary to transition from the pioneering phase to a more structured and systematic approach, similar to the evolution seen in general software development with the advent of Platform Engineering movement, to enhance operational efficiency, reduce costs, and streamline processes in ML/AI operations.

Amazon EKS, besides being one of the most used Kubernetes distributions and managed Kubernetes service worldwide, brings several limitations: in this blog post, we discover how Kamaji can overcome these, maintaining the same ease of a managed service thanks to its Auto-Pilot feature.

In the early days of Kubernetes adoption, single-cluster deployments were the norm, offering a straightforward approach to managing applications and services. As the adoption of Kubernetes expanded, the limitations of single-cluster models surfaced. The increasing demand for Kubernetes clusters requires a shift to multicluster deployments and an innovative Hosted Control Plane architecture.

As organizations continue to navigate the complex landscape of container orchestration, Kubernetes stands out as a leading solution for deploying, managing, and scaling containerized applications.However, on-premises infrastructures with limited availability zones face unique challenges, particularly when it comes to ensuring the resilience of the etcd cluster – a critical component for maintaining the state of the Kubernetes cluster.

Kubernetes on bare metal maximizes performance by eliminating virtualization overhead, providing direct access to physical hardware resources. This minimizes latency, optimizes resource utilization, and ensures consistent, high-speed communication between containers, making it an ideal choice for performance-critical workloads.

Kubernetes at the edge extends the powerful container orchestration capabilities of Kubernetes to decentralized environments. It enables the efficient management and deployment of containerized applications on edge devices, reducing latency, enhancing scalability, and supporting real-time processing closer to data sources.

Open-source software has become the backbone of modern technology, powering everything from operating systems to Web browsers and applications. Its collaborative nature fosters innovation and enables a wide range of solutions to common problems. However, as the popularity of open-source projects grows, so does the need for robust support and security measures. In this article we will explain how offering a commercial-grade solution to an open-source project such as Capsule is not only consistent with the project but also essential to its success.

Kubernetes is a powerful tool for container orchestration, but managing its control plane can be challenging, especially when running multiple clusters on different infrastructures and cloud providers. In this article, we'll explore Kamaji, Clastix's innovative architecture for simplifying control plane management and how it can make running multiple Kubernetes clusters on any infrastructure a breeze.

In this interesting article, we delve into the current law & enforcement applications of K8s and hypothesize even more interesting uses that have barely been tested in the defense field.

CLASTIX is working on evangelizing the benefit of addressing multi-tenancy on Kubernetes in the right path, as we’re working on drafting the first-ever State of the Multi-Tenancy report.So far, we had the chance to talk to many companies that are running up to thousands of containers on Kubernetes and going to solve their governance issues with multi-tenancy: however, let’s try to understand what is multi-tenancy, first.

Open source technology is essential for European countries to achieve data sovereignty and independence from American and Chinese regulations, while fostering innovation and providing greater security and privacy over data.

CLASTIX is now a SUSE ecosystem partner. Customers can now take advantage of innovative solution that benefits from SUSE’s decades of engineering excellence and open source leadership coupled with CLASTIX’s unique multi-tenant architecture for Kubernetes.

In this post, we will show you step by step how to install Capsule on KIND to bring multitenancy to your local cluster.

Discover all the great new features of the Capsule V0.1.2 release!

At CLASTIX, we're proud to announce Kamaji a new open-source tool to build and operate Managed Kubernetes Services with unprecedented ease.

As multi-tenancy continues to remain a hot topic in the market, we are proud to announce the next release of Capsule multi-tenancy operator v0.1.1!

With the accelerating adoption of Kubernetes the dilemma of multi-tenancy is gaining more traction and being more hot and relevant. Implementing a multi-tenancy architecture is proving to be one of the more strategic decisions that need to be considered when implementing Kubernetes-based infrastructure, as it has long term implications to operational efficiency, scale velocity and costs.