VMware's Kubernetes Evolution: Quashing Complexity

Key Concepts

• Kubernetes complexity
• Virtualization benefits and overhead
• Platform engineering
• VMware Cloud Foundation (VCF)
• VMware Kubernetes Service (VKS)
• Private AI
• GPU virtualization
• Open-source contributions
• Multi-cluster management
• V-Mug Community

Kubernetes Complexity and Platform Engineering

• Challenge: Kubernetes, while offering automation and desired state management, presents significant complexity.
• Statistic: 84% of organizations are evaluating or deploying Kubernetes (CNCF 2023 survey).
• Statistic: 46% of organizations lack training and find complexity the biggest challenge (CNCF 2023 survey).
• Complexity Factors:
  • Beyond the Kubernetes runtime, integration with networking, storage, backup, monitoring (Prometheus), registry (Harbor), and service mesh (Antrea/Envoy) is required.
  • Developers spend excessive time on infrastructure services instead of core application development.
• Solution: VMware Kubernetes Service (VKS) within VCF provides a pre-configured Kubernetes platform.
  • Includes controller runtime, Harbor registry, Velero backup, Envoy/Antrea network management, and CSIs for storage.
  • Platform engineering teams can define the desired spec for the platform and its services, reducing the burden of maintaining open-source components.
• Platform Engineering Role: Platform engineering is more effective when it can leverage a service like VKS to abstract away the underlying complexity.

VMware Cloud Foundation (VCF) as the "New vSphere"

• Core Functionality: VCF virtualizes compute, storage, networking, and associated services.
• Kubernetes Integration: Native Kubernetes runtime is built into the platform, along with essential services.
• IAS Delivery: VCF delivers IAS services directly to application users, including load balancing, network management, and firewall services.
• Administrator Experience: Simplifies operations for administrators familiar with vSphere, enabling them to manage both VMs and containers on a single platform.
• Maintenance: VMware maintains the open-source components, versions, upgrades, and support, reducing the operational burden.
• Policy-Based Configuration: Kubernetes deployment and service configuration are managed through policies.
  • Example: Defining a distributed Kubernetes control node with a minimum of three hosts.
  • Multi-tenancy support allows different application teams to have customized IAS layers with specific service sets.

Virtualization Overhead and Benefits

• Myth Debunked: Virtualization does not have a significant overhead.
• ML Perf Benchmarks: VMware's ML Perf benchmarks on Kubernetes show negligible performance drop with full virtualization.
  • Achieved 98.3% of bare metal performance on average.
• Benefits of Virtualization:
  • Dynamic application and pod movement across clusters.
  • Standardized way of running servers and services.
  • Non-disruptive maintenance and updates.
  • Mitigation of server failure rates.

Addressing Infrastructure Management Complexities

• Non-Disruptive Operations: Virtualization enables non-disruptive updates to systems, drivers, and other core infrastructure components.
• Service Integration: VCF simplifies the integration of essential Kubernetes services, reducing the complexity of managing multiple open-source modules.
• Kubernetes' Core Value: Kubernetes enables consistent and scalable application deployment by defining infrastructure state.
• Complexity Removal: VMware focuses on removing complexity and simplifying infrastructure management for Kubernetes.
• Example from VCF 5.2: Simplified Harbor service deployment by enabling it directly within the cluster spec.
• Multi-Cluster Management: Tanzu Mission Control is now available to all VCF customers for centralized management of applications across multiple VCF instances and regions.

Kubernetes Version Management

• Challenge: Limited support duration for Kubernetes releases (approximately 1 year and 4 months) requires continuous updates.
• Solution: VCF supports multiple Kubernetes versions concurrently, allowing applications to use different validated versions.
• Support: VMware provides full support for all shipped components across multiple Kubernetes versions.

Private AI and Kubernetes

• Driving Force: AI is a significant driver of Kubernetes deployments.
• Containerization: AI applications are typically container-based and Kubernetes-based.
• Private AI Drivers: Data sovereignty and trustworthiness are driving the need for private AI solutions.
• Requirements for Private AI:
  • Platform support for AI applications.
  • High-performance storage with high streaming throughput.
  • High-bandwidth networking interfaces.
  • Full Kubernetes runtime with essential services.
• VMware's Role: VMware provides a platform for running AI services within an organization.
• Partnerships: VMware collaborates with Nvidia (NGC), the open-source community (OPIA), and UC Berkeley on AI initiatives.

VMware Private AI Foundation

• Purpose: Simplifies the deployment and management of AI applications in a private environment.
• Features:
  • Out-of-the-box support for Nvidia GPU Cloud (NGC) services.
  • Extended registry service for managing open-source models.
  • Pre-configured application runtimes.
• GPU Virtualization: Enables higher utilization of expensive GPU resources through virtualization.
  • VMware testing shows non-disruptive vMotion of AI applications with a 70 billion parameter Llama model.
  • Achieves similar or better performance than bare metal with GPU virtualization.
• Pre-Configuration: Ships with AI-ready virtual machines that include GPU drivers and necessary services.

Open Source Contributions

• Commitment: VMware is a top contributor to Kubernetes and a member of CNCF.
• Key Projects: VMware leads or actively participates in projects like Harbor, Antrea, Envoy, Cluster API, Valero, and Contour.
• Vision: VMware aims to ensure that Kubernetes remains an open, standards-based application layer that is cloud-independent.

Community Engagement and Learning Resources

• V-Mug Community: A 120,000-member community worldwide offering meetups and learning events.
• VMware Explore: Annual conference in Las Vegas.
• Online Resources: Webinars and documentation available on vmware.com.
• Free Training: VCF customers receive free access to all training and services.
• Free Licenses: Developers can obtain free licenses for personal use and home labs.

Conclusion

VMware is focused on simplifying the complexities of Kubernetes and enabling efficient resource utilization, particularly in the context of AI. VMware Cloud Foundation (VCF) and VMware Kubernetes Service (VKS) provide a platform that abstracts away the underlying infrastructure complexities, allowing organizations to focus on application development and service delivery. VMware's commitment to open source and community engagement further supports the adoption and innovation of Kubernetes and AI technologies.