Docs

Volcengine Kubernetes Engine

Overview

Application scenarios

This topic describes the common application scenarios of Volcengine Kubernetes Engine (VKE).

Distributed inference

The prefill-decode (PD) disaggregation deployment improves resource management efficiency. PD instances can be dynamically scaled out to enhance GPU utilization.

AI training

VKE enables distributed scheduling and management of AI training tasks and supports dynamic allocation of multi-node GPU resources and elastic scaling.

Reinforcement learning

Flexible multi-role orchestration and elastic high-concurrency scheduling improve training efficiency. The diagnostics and self-healing capabilities of clusters ensure training stability.

Highly elastic workloads

VKE provides stable, reliable, on-demand resource support for workloads with highly volatile traffic, such as live streaming, e-commerce, and gaming, and can scale resources in seconds.

Traffic-based autoscaling or self-healing helps keep services running smoothly and stably.

Seamless integration with Volcengine Container Instance (VCI) enables you to launch large numbers of instances in seconds and flexibly handle traffic peaks.

The use of VCI enables you to focus on building your applications without the need to purchase and manage the infrastructure, such as underlying servers.

Separation of applications into microservices

Traditional monolithic applications can be refactored into containerized microservices to improve the agility, maintainability, and scalability of business modules.

Canary deployment, autoscaling, and end-to-end data monitoring are supported to improve enterprise productivity.

Integrated microservice support, together with container capabilities such as rapid updates, rapid iteration, and rapid deployment, helps enterprises implement microservices smoothly in production environments.

Agile development for accelerated iteration

Code compilation, image building, canary deployment, and application deployment are automated based on code sources.

A container image service provides the foundation for DevOps, enabling automated code compilation, image building, canary deployment, and application deployment from source code, and supporting end-to-end delivery of containerized applications.

The entire DevOps workflow, from code changes and image building to application delivery, deployment, and O&M, can be automated. This automation helps teams release new features faster, reduce differences between production and test environments, and accelerate iteration.

Multi-cloud deployment and interconnection

Based on the multi-cloud active-active architecture, private cloud and public cloud environments are connected and managed in a centralized manner. This implements the seamless migration of applications and data between cloud and on-premises infrastructure.

Multiple heterogeneous cloud resources can be operated, maintained, and managed to achieve cross-cloud resource scheduling and flexible disaster recovery for businesses.

Last updated: 2026.07.16 15:55:45