KubeCon + CloudNativeCon India 2025

In the world of containerization, workloads require reliable, performant storage for logs, data persistence, debugging, and real-time processing. Today, Kubernetes and open-source solutions provide a diverse range of storage provisioning options to choose from for single-cluster environments.

Now, envision scaling this to a broader fleet model, where multiple clusters are in play—worker clusters managed by a single control cluster, supporting a variety of workloads with replica counts in the range of thousands. How can we ensure seamless, dynamic volume provisioning through intelligent, capacity-aware decisions, while optimizing disk utilization across clusters, minimizing manual intervention, and maintaining high performance, reliability, and security as workloads and customer demands grow?

In this session, we will tackle these challenges head-on, exploring the architecture of capacity-aware, dynamic volume provisioning for LVM local storage across a multi-cluster Kubernetes fleet.

Running large-scale, performance-sensitive distributed databases on K8s requires balancing between latency, reliability, and cost efficiency. At Flipkart, we tackled this challenge by leveraging K8s Local Persistent Volumes (Local PVs) to optimize both performance and infrastructure costs for our stateful workloads.

With a storage footprint of 4PB of directly attached SSDs, 15PB of directly attached HDDs, comprising 1000 K8s stateful nodes spanning across 3 clusters (2 on-prem & 1 on public cloud infrastructure) LocalPV storage currently powers various internal database platforms (Aerospikes, HBase, TiDB) leveraged by critical business applications at Flipkart.

In this talk, we’ll share practical insights from our journey of scaling database infrastructure with Local PVs. We’ll dive into key challenges and solutions, including dynamic provisioning, workload scheduling, and automating maintenance workflows to ensure operational efficiency at scale.

In cloud-native environments, high-performance networking is crucial for scalability and security. Calico, a leading solution for container networking, integrates the VPP dataplane alongside eBPF. VPP provides a scalable solution combining security with high-performance packet processing. In this session, we’ll discuss our work with Calico’s VPP dataplane, optimized by integrating Marvell’s high-performance native plugin, an open-source Data Acceleration Offload solution. This integration offloads cryptographic tasks and inline IPSec encryption to SmartNICs, reducing CPU usage and optimizing secure data transmission. VPP offers fast packet processing, low latency, and high throughput, essential for service meshes, edge computing, and microservices. By leveraging hardware offloading and optimizations, VPP enhances throughput and reduces CPU load, making it ideal for demanding networking tasks. We’ll present benchmark results showcasing improved throughput, latency, and resource use.

This session delves into the critical aspects of developing production-ready Large Language Model (LLM) applications using Java. We'll explore how to leverage Java's strengths to build scalable and efficient LLM systems, addressing key challenges such as performance optimization, resource management, and seamless integration with existing infrastructures.

Attendees will gain practical knowledge on handling massive datasets, optimizing model inference, and fine-tuning LLMs for optimal performance. We'll discuss strategies for ensuring the reliability and scalability of your LLM deployments, empowering you to create robust and high-performing AI applications. Whether you're a seasoned Java developer or new to the AI domain, this session will provide valuable insights and guidance for your LLM development journey, equipping you with the tools and knowledge to navigate the complexities of building production-grade LLM systems.

Kubernetes Operators are revolutionasing cloud-native application management—but when designing an operator, should we go one CRD per operator or manage multiple CRDs within a single operator? How does cascading CRDs impact reconciliation, dependency handling, and performance?

In this talk, we'll dive deep into real-world use cases to explore the trade-offs between these approaches. Whether we are building custom Kubernetes controllers or optimising cloud-native infrastructure, this session will help one make the right design choice for scalability, maintainability, and performance.

Kubernetes-native development is exhilarating—until complex workflows, slow feedback loops, and steep learning curves slow you down. Teams often spend days or weeks setting up environments, managing dependencies, and debugging microservices. GitLab found that nearly 44% of organizations take over two months to onboard new developers—precious time lost in today’s fast-paced tech world.
This session shows how we cut environment setup from days to minutes, slashed feedback loops by over 90%, and reduced onboarding time by 50%.
The secret? A powerful blend of CNCF tools—Kind and DevSpace—combined with simple lightweight Node.js and Bash/PowerShell scripts, We have automated environment provisioning, code synchronization, and cluster management so that developers can focus on shipping features.
Discover how to integrate Kind and DevSpace for a seamless workflow with real metrics and tips. Developers, architects, and DevOps can streamline Kubernetes, speed onboarding, and boost iterations.