kmesh

Introduction

Kmesh is a high-performance and low overhead service mesh data plane based on eBPF and programmable kernel. Kmesh brings traffic management, security and monitoring to service communication without needing application code changes. It is natively sidecarless, zero intrusion and without adding any resource cost to application container.

Why Kmesh

Challenges of the Service Mesh Data Plane

Service mesh software represented by Istio has gradually become popular and become an important component of cloud native infrastructure. However, there are still some challenges faced:

• Extra latency overhead at the proxy layer: Add 2~3ms latency, which cannot meet the SLA requirements of latency-sensitive applications. Although the community has come up with a variety of optimizations, the overhead introduced by sidecar cannot be completely reduced.
• High resources occupation: Occupy 0.5 vCPU and 50 MB memory per 1000 requests per second going through the proxy, and the deployment density of service container decreases.

Kmesh Architecture

Kmesh transparently intercepts and forwards traffic based on node local eBPF without introducing extra connection hops, both the latency and resource overhead are negligible.

The main components of Kmesh include:

• Kmesh-daemon: The management component per node responsible for bpf prog management, xDS configuration subscribe, observability, and etc.
• eBPF Orchestration: The traffic orchestration implemented based on eBPF, supports L4 load balancing, traffic encryption, monitoring and simple L7 dynamic routing.
• Waypoint: Responsible for advanced L7 traffic governance, can be deployed separately per namespace, per service.

Kmesh innovatively sinks Layer 4 and Simple Layer 7 (HTTP) traffic governance to the kernel, and build a transparent sidecarless service mesh without passing through the proxy layer on the data path. We named this Kernel-Native mode.

Kmesh also provides a Dual-Engine Mode, which makes use of eBPF and waypoint to process L4 and L7 traffic separately, thus allow you to adopt Kmesh incrementally, enabling a smooth transition from no mesh, to a secure L4, to full L7 processing.

Key features of Kmesh

Smooth Compatibility

• Application-transparent Traffic Management

High Performance

• Forwarding delay 60%↓
• Workload startup performance 40%↑

Low Resource Overhead

• ServiceMesh data plane overhead 70%↓

Zero Trust

• Provide zero trust security with default mutual TLS
• Policy enforcement both in eBPF and waypoints

Safety Isolation

• eBPF Virtual machine security
• Cgroup level orchestration isolation

Open Ecology

• Supports XDS protocol standards
• Support Gateway API

Quick Start

Please refer to quick start and user guide to try Kmesh quickly.

Guided Install

KubeStellar Console provides a guided installation mission for Kmesh with step-by-step instructions, pre-flight checks, validation, troubleshooting, and rollback support. It works against your live cluster via kubeconfig, or as read-only documentation without a cluster connection.

Performance

Based on Fortio, the performance of Kmesh and Envoy was tested. The test results are as follows:

For a complete performance test result, please refer to Kmesh Performance Test.

Contact

If you have any question, feel free to reach out to us in the following ways:

• meeting notes
• mailing list
• slack
• twitter

Community Meeting

Regular Community Meeting:

• Thursday at 16:00 UTC+8 (Chinese)(weekly). Convert to your timezone.

Resources:

• Meeting Link

Contributing

If you're interested in being a contributor and want to get involved in developing Kmesh, please see CONTRIBUTING for more details on submitting patches and the contribution workflow.

License

The Kmesh user space components are licensed under the Apache License, Version 2.0. The BPF code templates, ko(kernel module) and mesh data accelerate are dual-licensed under the General Public License, Version 2.0 (only) and the 2-Clause BSD License (you can use the terms of either license, at your option).

Credit

This project was initially incubated in the openEuler community, thanks openEuler Community for the help on promoting this project in early days.