The Sched app allows you to build your schedule but is not a substitute for your event registration. You must be registered for KubeCon + CloudNativeCon Europe 2025 to participate in the sessions. If you have not registered but would like to join us, please go to the event registration page to purchase a registration.
Please note: This schedule is automatically displayed in British Summer Time (BST) (UTC +1). To see the schedule in your preferred timezone, please select from the drop-down menu to the right, above "Filter by Date." The schedule is subject to change and session seating is available on a first-come, first-served basis.
Sign up or log in to bookmark your favorites and sync them to your phone or calendar.
Join us for a rapid-fire journey through the CNCF ecosystem, where experts, including project maintainers and community members, share insights, innovations, and real-world applications of Cloud Native Computing Foundation projects. Each project has just five minutes to present, promising to enlighten and inspire with cutting-edge tools and practices that shape the future of cloud-native development. Whether you're a seasoned pro or just getting started, there's something for everyone in the world of Cloud Native Computing!
FAQ: Do I need an all-access pass to attend the project lightning talks? No, you will only need your KubeCon + CloudNativeCon only pass for access.
“Get your hiking boots ready because we are about to traverse the wild, wonderful world of the CNCF Landscape. Why you ask? We currently have over 190 projects, and finding information about them can be a challenge. “Just go to the website” isn’t enough, sometimes you need a guide to show you the ropes. In these introductory sessions we will go over some of the diverse set of projects inside the CNCF so that you’re well equipped to find what you’re looking for at KubeCon.
Join Taylor for the fastest review of the CNCF Wasm Ecosystem you've ever seen. This talk will give you the (literal) 30 second overview of Wasm before a rapid fire review of the projects and advancements in the CNCF Wasm ecosystem. This will be your super fast, one-stop shop for all updates Wasm.
The Common Expressions Language (CEL) is a fast, portable, non-Turing-complete evaluation language that can simplify writing and reusing policies, and is finally here since Kubernetes 1.29: from validating CRDs to ValidatingAdmissionPolicies.
In this lightning talk we will showcase how Kubewarden, thanks to WebAssembly, reuses the very same CEL libraries from Kubernetes to provide a CEL policy that is backwards compatible. We will also learn how this Kubewarden CEL policy extends the CEL features by providing context-aware capabilities, telemetry and policy reports.
This lightning talk will provide you a comprehensive overview about the latest project development efforts around the Kubernetes container runtime CRI-O. We'll discuss how the project influences the runtime space, Kubernetes development as well as what we plan for the future.
Currently, the RDMA device plugin cannot coordinate the allocation of PCIe-affined RDMA devices based on GPU device allocation and requires users to manually ensure consistency between RDMA resources and Multus network configurations in the Pod YAML. To address this, Spiderpool introduces DRA to declare RDMA resources and network interfaces, significantly simplifying the configuration complexity in Pod YAML. Additionally, during Pod startup, Spiderpool leverages NRI to dynamically insert RDMA devices with GPU affinity, ensuring optimal GPU Direct RDMA performance.
Furthermore, existing community projects lack support for exporting metrics for isolated RDMA devices based on SR-IOV. Spiderpool bridges this gap by exporting comprehensive RDMA metrics for both Pods and nodes to Prometheus. It enables RDMA observability from multiple perspectives, including AI tasks.
These features represent groundbreaking innovations within the community, effectively empowering AI workloads.
Kuma is an open source control plane that delivers advanced traffic management and observability for microservices. In this session, we’ll talk about the latest releases and most exciting features from each.
Multi-platform images are commonly used in IoT and Edge computing, particularly heterogeneous deployments, as well as cross-platform environments. There are a few tools that enable users to create multi-platform images, such as docker, buildx, podman, etc. However, there are limitations in some network-constrained scenarios, such as IoT and Edge, and air-gapped environments. These tools have to rely on remote registry to create multi-platform images and lack of native support for local dev/testing.
ORAS is a registry client tool that helps you not just create a multi-platform image and OCI artifact, but also enables you to associate supply chain artifacts (e.g. signature, SBoM, vulnerability report) with your multi-platform images for secure distribution. ORAS supports creating and managing a multi-platform image in both local environment and OCI registries.
In this session, Feynman will demonstrate how to create a multi-platform image and distribute it with security guarantee.
The open-source ecosystem powers the foundations of modern technology, yet its security challenges remain a critical concern. Vulnerabilities in open-source projects can ripple across industries, affecting millions. Security engineers possess the expertise to identify, mitigate, and prevent these risks.
This presentation highlights the vital role security engineers can play in the community, not only in fortifying individual projects but in raising the security ceiling for the entire ecosystem. Attendees will learn how contributing to open-source security efforts can amplify their impact, improve their skills, and build a more resilient landscape. From threat modeling, vulnerability disclosure, and tooling development, security engineers can influence the security posture of projects used worldwide.
Join us to discover how your expertise can make a difference, gain insights into impactful contribution strategies, and become part of a movement that redefines open-source security.
Software distribution encompasses not only images but also binaries, which are often distributed outside of OCI registries, such as through websites or package managers. It is essential for teams to produce SBOMs for these binaries and distribute them together. Ensuring the security of these binaries is as critical as securing regular OCI artifacts. While existing tools like GPG can perform basic signing tasks, they lack extensibility and do not offer fine-grained signature verification.
Notation from the Notary Project addresses these limitations by enabling the signing of arbitrary blobs beyond regular OCI artifacts. It also provides verification based on fine-grained trust policies, thereby enhancing security. Furthermore, Notation's plugin model allows for flexible and robust security measures tailored to specific needs.
In this session, Shiwei will demonstrate how Notation performs blob signing and verification, showcasing its application in securing binary releases of software.
This talk's goal is to bring people up to date with everything that happened on the Kairos project the last 12 months, especially new features and user facing functionality. It will also be useful to people not already familiar with the project, to see what it does and how active the project is.
Metal3 has come a long way since its humble beginning and is now targeting incubation. This talk will give a brief look back at what we have achieved in the sandbox phase, and then focus on how we are now maturing and evolving the project going forward.
K3s is a lightweight, production-ready Kubernetes distribution perfect for edge computing, IoT devices, and CI/CD. It's known for its small footprint, simple installation, and CNCF conformance.
This lightning talk will quickly get you up to speed on the latest and greatest in K3s:
K3s in a Nutshell: We'll start with a brief overview of K3s, highlighting what makes it unique in the Kubernetes landscape.
Recent Developments: Explore key changes and improvements from recent releases.
Roadmap: A glimpse into the future of K3s, including planned features and development priorities:
Want to Get Involved?: We'll wrap up with tips for new contributors, showing you how to engage with the team, share your ideas and get ready for your first contribution
In today’s cloud-native environments, security is paramount, and a "Zero Trust" approach is crucial to ensuring sensitive data remains protected. This lightning talk will demonstrate how to implement a Zero Trust model for secrets management in Kubernetes using the External Secrets Operator (ESO). ESO seamlessly integrates external secret stores with Kubernetes, enabling secure, automated synchronization and distribution of secrets across multiple clusters and cloud providers. We will explore how ESO can source secrets from trusted, external systems with fine-grained access controls. Learn how to ensure only the necessary services can access sensitive data while maintaining complete transparency and control over your secrets management workflow. In just five minutes, you’ll walk away with practical insights into securing your Kubernetes clusters with ESO, and how a Zero Trust approach minimizes attack surfaces while maximizing operational efficiency.
Attestations consist of authenticated statements about a software artifact or a collection of artifacts, as defined by the SLSA attestation. Examples include signed provenance files or signed SBOM files for container images. Attestations are vital for ensuring the integrity and trustworthiness of the software supply chain for container images.
Ratify, a CNCF sandbox project, provides a comprehensive framework for verifying artifact security metadata, such as signatures and attestations, to ensure artifacts are trustworthy and compliant before they are used.
In this lightning talk, Yi Zha will give an overview of the Ratify project and attestations, and a demo showcasing using Ratify for securing K8s deployments through attestation verification. Attendees will gain valuable insights into improving their Kubernetes security posture by leveraging Ratify's capabilities.
This lightning talk will cover highlights from Istio's recent releases, updates, features, and other project related news for our fellow maintainers, contributors and users. Learn how our latest releases signals ambient mode – service mesh without sidecars – is ready for everyone, while we continue to support sidecars as first-class citizens too.
Join us on that lightning talk to see how we are performing with our projects under the LFX Mentorship program and why we think we MUST talk and evangelise more about the program!
We will use our two projects as examples - Harbor Satellite and Harbor CLI.
Almost every CNCF project dealing with Kubernetes ends up with CRDs that define the project’s API surface… which means that almost every CNCF project has to learn the same hard lessons about managing such a CRD API! Join us for a whirlwind look at the most crucial things that we’ve learned (usually the hard way!) creating and growing Linkerd, with special emphasis on the effect of shared CRDs like Gateway API on projects that need to work together using them.
Crossplane (https://www.crossplane.io/) has significantly matured over the 6 years of the project. It is full of powerful and expressive functionality that lets Platform Engineers build the platform of their dreams and customize it for their specific organization’s needs.
With all this power and flexibility, there are many different paths that can be taken to build your production ready control plane. In this talk, we will cover some high level guidance and best practices, as well as specific examples and tips, to help you streamline your Crossplane journey and get to production smoothly and rapidly.
We’ll share the patterns we’ve uncovered over years of building production platforms with Crossplane, and help you avoid common obstacles that we’ve helped our community overcome on their journeys as well. We’ll race against the clock to help you get to production in this fast paced Crossplane lightning talk!
Karmada (Kubernetes Armada) is a Kubernetes management system that enables you to run your cloud-native applications across multiple Kubernetes clusters and clouds.
In this presentation, the maintainer of the Karmada project will share: - A Brief introduction to Karmada - Key features and real-world use cases - New features over the last year - Next plan
KubeVirt and Containerized Data Importer (CDI) have been ported to ARM and s390x in addition to x86 cluster node architectures. Running these operators on a multi-arch Kubernetes cluster opens up a number of exciting use cases to import, manage, launch and template virtual machines. For example, while only x86 nodes have API access to a certain external commercial VM management platform, by importing a VM through the CDI operator deployed on an x86 node, the VM can be disassembled, templated and re-built as either a VM or a container to run on the s390x or ARM nodes, and tested all in the same K8s cluster. This makes KubeVirt and CDI particularly useful for modernizing complex workloads.
Open Cluster Management (OCM) addresses the challenges of managing multiple Kubernetes clusters, providing open APIs for cluster registration, workload distribution, dynamic placement of policies, and more. The placement concept allows dynamic selection of clusters, enabling users to replicate Kubernetes resources or run advanced workloads across member clusters. For instance, as an application developer, I can deploy workloads to clusters with the most available memory and CPU. With the rise of AI technology, there's an increasing need to schedule AI workloads based on GPU/TPU resources. In this talk, we will demonstrate how to utilize the extensible placement scheduling mechanism and a GPU/TPU resource collector addon. Using an addon template, this setup can provide an AddonPlacementScore, facilitating placement decisions based on GPU/TPU resources. This approach enables OCM API consumers to intelligently schedule AI workloads to the most optimal clusters.
Kubernetes ecosystem is rich yet complex, requiring engineers to handle intricate configurations across multiple components and custom resources. This complexity can be overwhelming, especially with tedious configuration management tasks like managing YAML files, network setups, and RBAC rules.
Meshery, an open-source cloud-native manager, addresses this by leveraging human-computer interaction principles to simplify Kubernetes. Through Kanvas, an intuitive visual interface, Meshery reduces cognitive load, aligns with users mental models, and streamlines infrastructure design backed by OPA policies. Join this lighting talk to see how Meshery makes Kubernetes more accessible, empowering you to visualize and manage complex relationships across CNCF projects with ease and precision.
Project Capsule has gained widespread adoption across organizations, enhancing Kubernetes platforms with robust Multi-Tenancy capabilities. In this session, we’ll showcase the project’s journey so far, emphasizing its evolution into a comprehensive toolkit for building Multi-Tenant Kubernetes platforms and its seamless integration with other CNCF projects.
k8gb is the Kubernetes-native, open source CNCF Global Service Load Balancing solution, enabling global traffic balancing via DNS. It has no dedicated management cluster and no single point of failure.
Interested in multi-cluster failover or geolocation-aware routing but never heard of k8gb? We'll have an overview.
Want to know what we've been up to? We'll tell you. Want to know what's coming next? We'll show you! Discover how k8gb is shaping the future of reliable, scalable, and open-source global balancing in Kubernetes.
Federated Learning (FL) enables collaborative model training while ensuring data privacy, a crucial requirement for many organizations. Open-Cluster-Management (OCM) extends this capability by managing both public and private clusters, making it an ideal solution for environments with strict data governance, such as private clouds. Instead of transferring sensitive data between clusters, OCM leverages FL to move models, significantly reducing bandwidth usage and minimizing the need for large-scale data storage within individual clusters.
Furthermore, OCM standardizes FL workflows, providing seamless integration with platforms like Flower, OpenFL, and FATE through a unified interface. This session will demonstrate how OCM enhances FL with scalable multi-cluster management, cost-efficient operations, and standardized workflows to enable the development of smarter, privacy-focused AI solutions.
Service mesh has brought transparent encryption capabilities to cloud-native applications. However, the tight coupling of sidecars with workloads complicates lifecycle management.
Both Isito and Cilium have implemented sidecarless service mesh, utilizing userspace proxies that increase connection hops and introduce single points of failure, with encryption occurring only between proxy links. In this session, we aim to demonstrate how eBPF and programmable kernel modules can significantly address these challenges.
We believe that offloading traffic to eBPF and leveraging kernel innovations to achieve end-to-end secure encryption capabilities is the future direction for sidecarless service mesh. We will discuss how this approach can provide a more efficient and secure network architecture without the need for sidecar proxies.
Finally, we will present use cases and discuss how to maintain encryption capabilities and minimize the impact on applications during scenarios such as failures.
Deploying and managing Kubeflow has traditionally been a complex undertaking. This session introduces a new, community-driven Helm Chart designed to simplify Kubeflow installation and management and make it more accessible to users of all levels.
This Helm Chart embraces a minimalist and modular design, focusing solely on deploying Kubeflow core components and integration manifests. This approach reduces complexity by leveraging existing infrastructure and pre-installed dependencies and empowers platform administrators to tailor deployments to their specific needs.
We'll explore the motivations behind this new Helm Chart and discuss how it can streamline Kubeflow deployments in diverse environments, from single-user setups to large, multi-team organizations. Finally, we'll highlight how this open-source project aims to benefit the wider Kubeflow community.
“Get your hiking boots ready because we are about to traverse the wild, wonderful world of the CNCF Landscape. Why you ask? We currently have over 190 projects, and finding information about them can be a challenge. “Just go to the website” isn’t enough, sometimes you need a guide to show you the ropes. In these introductory sessions we will go over some of the diverse set of projects inside the CNCF so that you’re well equipped to find what you’re looking for at KubeCon.
Maintaining secure container images and addressing new vulnerabilities quickly is a major challenge. To patch images, end users face two options: wait for third-party authors to release updates, which can take weeks, or perform a full image rebuild, a time and resource-intensive process.
Project Copacetic (Copa) enhances the image patching process, reducing turnaround time and complexity. It integrates easily into existing build infrastructure, giving users greater control over their patching timeline and reducing costs.
Copa scans container images using tools like Trivy to generate a vulnerability report and parses the report for necessary OS-level package updates. It applies these updates to the target image using Buildkit (Docker’s default builder) to create a new patch layer on the original image. Copa can even patch distroless images by leveraging external tooling.
The talk will overview Copa, highlighting new features and future improvements.
Managing and protecting stateful applications on Kubernetes is complex, particularly with interconnected workloads, databases, volumes, networking, and configuration resources. A robust data protection tool must address these challenges while ensuring resilience.
Kanister, a CNCF Sandbox project, enables users and adopters to create data protection workflows in Kubernetes using Custom Resources. By leveraging Kubernetes-native constructs, it provides seamless backup and disaster recovery for applications with persistent storage, including databases and AI/ML workloads.
In this lightning talk, Pavan will introduce Kanister, highlighting its capabilities and ongoing developments, such as support for vector databases, performance optimizations, and security enhancements. These updates ensure Kanister remains a cutting-edge solution for modern cloud-native applications.
Kubernetes SIG Storage is responsible for ensuring that different types of file and block storage are available wherever a container is scheduled. There is the Container Storage Interface (CSI) for block and file storage that allows storage providers to write CSI drivers. There is also Container Object Storage Interface (COSI) that is adding object storage support in Kubernetes. In this lightning talk, we will highlight some projects that SIG Storage has been working on.
Prometheus 3.0 was released on stage during KubeCon Salt Lake City. Join us for the most important 3.0 changes, for the newest updates since then, and for how we intend to 10x the maintainer base of Prometheus with the help of our new governance.
Falco, the CNCF runtime security project, can continuously monitor your entire environment looking for suspicious activity and do much more than that: thanks to its rich ecosystem of plugins and community projects it can visualize, forward and respond to security sensitive events. Join us in this lightning talk to learn about how easy it is to monitor and secure your cloud-native infrastructure with Falco and which performance and feature improvements we have been working on.
Learn about LitmusChaos, a CNCF Incubating project empowering teams to adopt chaos engineering for building resilient cloud-native systems. This session will provide an introduction to the project, its core principles, and how it enables controlled failure injection to uncover system weaknesses.
We’ll also focus on the ongoing efforts toward CNCF Graduation, including advancements in governance, security, and community engagement. Discover what this milestone means for the project and the open-source ecosystem, and how you can contribute to its journey.
Join us to explore the essence of LitmusChaos and its roadmap to becoming a CNCF Graduated project.
Update about the Perses project since last PromCon (2024). Status regarding the CNCF, features implemented (on three different aspects: GitOps, Embedded and the Perses application)
In this talk, we will demonstrate how Kubescape ApplicationProfiles provide a structured view of your application’s behavior as Kubernetes objects.
Application profiles capture critical runtime attributes, including process information, system calls, network activity, and file access patterns using eBPF.
Through a short technical hands-on, we will examine how application profiles are generated, what data they collect, and how users can leverage them to detect and enforce workload security.
Our update session will discuss the latest improvements and capabilities in the Kepler Project, a CNCF Sandbox power exporter crafted for Kubernetes. Kepler revolutionizes the estimation of container and pod power usage and offers insight into energy usage, efficiency, and carbon footprint. Some key breakthroughs in Kepler include power utilization features on Baremetal, the introduction of GPU metrics, validation tools for checking the accuracy of Kepler, progress with power models tailored for Virtual Machine environments, performance optimizations, and highlight in-progress features.
The session will also promise a unique opportunity to engage with one of the project's key maintainers, discuss the future of Kepler, contribute new ideas, and ask questions.
We all know that observability is a must-have for operating systems in production. But we often neglect our own backyard - our software release process. As a result, we also lack standardization, and each CI/CD tool invent its own way of reporting about pipeline runs, which causes fragmentation, lock-in and difficulty to leverage existing observability tools.
We've been talking about the need for a common "language" for reporting and observing CI/CD pipelines for years, and finally, we see the first "words" of this language entering the "dictionary" of observability - the OpenTelemetry open specification and semantic conventions.
On this lightning talk the CI/CD Observability SIG leads will share the work of the SIG.
The cloud native community prides itself on being open, innovative, and collaborative. However, systemic barriers to diversity and inclusion persist—even here. The TAG Contributor Strategy's new BIPOC (Black, Indigenous, and People of Color) initiative aims to address these challenges and foster a more equitable ecosystem.
In this lightning talk, we'll explore the "what," "why," and "how" of this initiative. What is the BIPOC initiative? A collective dedicated to empowering underrepresented voices in the cloud native space. Why does this matter? Despite our community's inclusive aspirations, barriers to participation remain, limiting the perspectives and contributions that drive innovation. How are we making a difference? Through advocacy efforts, building a supportive network, and increasing visibility for BIPOC professionals. Join us to learn about the tangible steps we're taking to create an inclusive space and the role you play in creating a more diverse cloud native future.
Recent years have seen a proliferation of large language models (LLMs) that extend beyond traditional language tasks to generative AI. This includes models like ChatGPT and Stable Diffusion. As this generative AI focus continues to grow, there is a rising need for a cloud native infrastructure that
Provides solid and scalable multi-cluster machine learning platform.
This talks will explore how these rising needs are addressed by leverage Volcano and Karmada that enable multi-cluster job queuing, management, enhanced scheduling.
This talk will cover: - The challenges for LLM training in single Kubernetes cluster\ - How to combine Volcano and Karmada to build a multi-cluster training platform - How to handle to job queuing across multi-cluster to ensure the fairness and SLA - How to balance the workload performance and multi-cluster utilization. - Scheduling policies to avoid busy waiting and dead lock
Vitess is a massively scalable horizontal distributed database system built for MySQL. This session will provide a brief overview of the project followed by what makes Vitess so scalable. It will conclude with data and charts to demonstrate these scalability claims.
Are you interested in building distributed applications or microservices, but don't know where to start? Join this session and learn how the Dapr building block APIs can make your life easier!
Dapr, the Distributed Application Runtime, provides a set of common APIs that makes building microservices a breeze. As a graduated CNCF project, Dapr is used in production by companies like Grafana, IBM, Alibaba Cloud & Microsoft and is a trusted OSS technology backed by a vibrant developer community.
In this lightning talk, I'll cover the most popular Dapr building block APIs and show how you can get started with it today!
For new users, we'll give a brief overview of how the graduated project, FluxCD, created GitOps. We'll also cover exciting updates with Flux v2.5 in 2025, which brings the Flux image automation APIs to general availability. We'll explain how image automation is core to updating a Git repository when new container images are available, and what GA means for you. Finally, we'll give some quick updates on security, integrations, and conformance testing for Kubernetes 1.32.
Cloud Native Buildpacks transform your application source code into images that can run on any cloud. They enable advanced caching mechanisms that improve performance at scale. They also allow for modularity and reuse, which ensure developers across your organization aren’t wasting cycles repeating what other teams have already done.
After this short talk, you’ll be able to run buildpacks with the Pack CLI and find off-the-shelf buildpacks in the Buildpack Registry, including those from Google, Heroku, and Paketo. Finally, you’ll learn how operators of large platforms use buildpacks to make their container builds as scalable as possible.
Strimzi, a CNCF incubating project, simplifies running Apache Kafka on Kubernetes with its operator-based approach. In this lightning talk, we'll provide an introduction to Strimzi, highlight key achievements from recent months, and share exciting plans for the future.
gRPC is a powerful framework for building high-performance, distributed systems. But with its growing ecosystem of tools and resources, it can be challenging to know where to start or how to deepen your expertise.
This talk is your guide to explore essential resources that will empower you to build, debug, and optimize your gRPC applications effectively.
This talk will be a short introduction to OpenFGA, a report on the progress of the project in the last few months, and an exploration of different adoption use cases from companies all sizes.
Developer Experience (DevX) is a key concept in Platform Engineering and in the cloud native ecosystem. Its primary goal is to empower developers, allowing them to focus on their code and generate business value rather than dealing with Kubernetes and infrastructure complexities.
Thanks to the Open Container Initiative (OCI) standard, the rise of cloud native runtimes has revolutionized how teams build and deploy applications. Developers leverage the same containerized approach from their local development workflows to remote environments. But this still exposes some inherent complexities. What is the right level of abstraction? How to reduce the cognitive load? How to shift down to the platform rather than shift left to the developers?
Let’s see in action how we answer these questions with Score, an open-source workload specification that enables developers to deploy their workloads across a spectrum of runtime platforms like Docker Compose, Kubernetes and more.
Porter packages your application, configuration, and scripts into an installer that is easy to distribute and use. Mixins are the special seasonings that help make Porter bundles. They are extensions of Porter that allow Porter to work with a tool or system like Helm, Terraform, or Docker. Together we’ll go over what Porter does, how it uses mixins, and how to make your own! Afterwards you’ll be equipped with the knowledge to contribute your own mixin to Porter.
What if AI could make modernizing legacy applications easier? Konveyor AI (Kai) is an open source project that automates migrations to cloud native technologies by combining static analysis with AI-driven agents. Its Reactive Code Planning Loop (RCPL) uses static analysis tools to detect migration issues, then generates tasks that agents can resolve: refactoring code, validating changes, and adapting new patterns.
With its bring-your-own-model design, Kai supports multiple LLMs, making it highly flexible and customizable. In this lightning talk, we’ll show how Kai’s integration of static analysis and AI-driven agents streamlines application modernization, while staying open, extensible and ready for cloud native-scale challenges.
Inference workloads are becoming increasingly prevalent and vital in Cloud Native world. However, it's not easy, one of the biggest challenges is large foundation model can not fit into a single node, which brings out the distributed inference with model parallelism, again, make serving inference workloads more complicated.
LeaderWorkerSet, aka. LWS, is a dedicated multi-host inference project aims to solve this problem, it's a project under the guidance of Kubernetes SIG-Apps and Serving Working Group. It offers a couple of features like dual-template for different types of Pods, fine-gained rolling update strategies, topology managements and all-or-nothing failure handlings.
What's more, vLLM, an inference engine, renowned for its performance and easy-to-use, has gained widespread popularity. In this presentation, we'll show you how to use LWS to deploy distributed inference with vLLM on Kubernetes.
