KubeCon + CloudNativeCon India 2024

Managing hybrid environments with public and private clouds poses significant challenges, especially with disparate interfaces and legacy applications. Our development and QE teams faced difficulties navigating multiple environments—testing private cloud code on private Docker clouds and public cloud code on public resources. Legacy apps, often running on bare metal or metal instances, added complexity. To tackle this, we used Kubernetes and CNCF projects like KubeVirt, CrossPlane, and Longhorn to build a unified hybrid hyper-converged cloud platform. This session will explore how we integrated these technologies to streamline operations. You'll learn about the initial challenges, how KubeVirt enabled virtualization, CrossPlane’s role in multi-cloud management, Longhorn’s storage solutions, and the unification of these components into a cohesive platform.

At Intuit, we have ~2000 production services that serve North-South and East-West Traffic using API Gateway and Service Mesh. These services span across ~350 k8s clusters and are managed by 10,000+ engineers. This talk walks you through multiple design iterations conducted by the Traffic team at Intuit in building the platforms that serve 1.5 Million+ TPS. Topics that are covered include lessons learnt while 1. Designing a cost effective, low latency, highly scalable, multi tenant API Gateway and Service Mesh platforms. 2.Using a non blocking model to externalise Authorization using an OPA based sidecar. 3.Distributing traffic configurations that are self serviced by 10,000+ engineers to the Traffic runtime platforms. 4. Designing Rate Limiting solutions with Low latency and low error

In Intuit, We have API Gateway and Service Mesh platforms that are responsible for handling ~30 billion requests every day and providing ingress to ~2000 services. With this talk, we’ll see - * How Admiral manages failover configurations by creating required custom resources like globalTrafficPolicy (GTP) across 300+ clusters and provides feedback on the activity. * Manage Observability for all failover activity and provide better insights to Developers. * How Intuit achieved simplified centralized failover UI experience leveraging Kubernetes and Istio features. * Operational challenges associated with managing the failover configurations. * High Availability and DR for services across 7000+ namespaces and enabling more than 900 teams to have seamless and unified experience to manage failover endpoints.

Applying modern deployment strategies, such as incremental releases, to complex ML architectures presents unique challenges, including managing resource constraints and ensuring cost efficiency. In the GPU environment, new replicas can take up to an hour to become operational due to extended image pull times. Despite these challenges, implementing advanced deployment strategies is crucial to minimise the impact of faulty versions, much like in traditional CPU workloads. In this session we will present on how to overcome these hurdles in GPU-based workloads. We will: - Start with traditional deployment approaches. - Progressively address advanced scenarios, such as implementing cost-efficient progressive delivery for high CPU-consuming workloads on GPU nodes and orchestrating sequential multi-resource deployments. - Reference our recent experience designing CI/CD for the GenAI applications. - Conclude by sharing cost comparison and best practices for GenAI model deployments.

How do we build a self-service platform that allows developers to interact with a legacy application through Kubernetes? In most other parts of the organisation, developers seek to operate on Kubernetes using the now-familiar concepts of GitOps and cloud-native ways for deploying their applications. When Kubernetes is a one-stop for deployment, the cognitive loads for platform and infrastructure can be minimized, and gives better focus to implementing business needs. In this talk, Bankdata.dk will share their experience building a Kubernetes API-driven self-service platform for operating three of their most critical legacy systems on Kubernetes, through ArgoCD and GitOps. The platform builds on the Java operator framework, and wraps operations on a portal server also running in Kubernetes. Bankdata.dk delivers all IT to 9 danish banks, including the 2nd and 3rd largest.

At PepsiCo, we are committed to enhancing efficiency and developer experience by continuously reviewing our processes and systems. As part of streamlining our deployment workflows, we transitioned to the GitOps methodology and adopted ArgoCD, a powerful GitOps tool with a feature-rich web UI that facilitates intuitive management of deployments, rollbacks, and application health. By leveraging ArgoCD's App of Apps pattern, we could define a bootstrapper Application, enabling the automation of application deployment. This approach allowed us to achieve self-service Continuous Delivery, empowering teams to independently manage their deployments while maintaining centralized control and visibility. Join us, as we share some of the challenges, including managing dependencies, configuring environments and ensuring consistency across deployments. Additionally, we will highlight the strategies and best practices, including the implementation of App of Apps pattern and the integration of Helm.