For kernel acceleration, the strongest matches are facebookincubator/katran (Katran is an eBPF-based Layer 4 load balancer that), dpdk/dpdk (This framework provides user-space network stacks and poll-mode drivers) and f-stack/f-stack (F-Stack is a DPDK-based kernel-bypass and user-space network stack). cilium/cilium and openvswitch/ovs round out the shortlist. Each is ranked by relevance to your query, popularity and recent activity.
Hand-picked open-source kernel acceleration software, ranked by GitHub stars and activity. Compare the top tools and find the best fit.
Katran is an eBPF-based Layer 4 load balancer designed for high-performance network packet forwarding directly within the Linux kernel. It utilizes an XDP network packet processor to bypass the standard kernel network stack, minimizing latency and maximizing throughput for incoming traffic. The system implements weighted consistent hashing to distribute network loads and maintain session persistence across backend servers. It further optimizes the return path through a direct server return gateway, which allows backends to respond directly to clients and removes the load balancer from the out
Katran is an eBPF-based Layer 4 load balancer that uses XDP for high-performance packet processing directly in the Linux kernel to achieve low latency and high throughput.
DPDK is a packet processing framework and kernel bypass networking library designed to accelerate network data plane performance. It provides a user space network stack that moves network data directly from the network interface card to user space, eliminating operating system kernel overhead to achieve high throughput and low latency. The framework utilizes a poll mode driver architecture that replaces interrupt-driven networking with a constant polling mechanism. This approach allows for the implementation of network functions and protocols outside the kernel. The project covers capabiliti
This framework provides user-space network stacks and poll-mode drivers for direct I/O and kernel bypass packet processing, perfectly matching the requirements for low-latency network acceleration.
f-stack is a kernel-bypass framework and user-space TCP/IP stack designed to increase packet processing speeds and overall throughput. It utilizes DPDK to route network traffic directly to user space and integrates a transplanted FreeBSD network stack to handle transport and network layer protocols outside of the host operating system. The project provides a socket API compatibility layer that allows standard network applications to run on a high-performance backend without requiring code changes. It also features a coroutine-based network API and micro-threading interface to manage asynchron
F-Stack is a DPDK-based kernel-bypass and user-space network stack designed for high-performance packet processing, missing only native eBPF support while squarely fulfilling the kernel acceleration requirements.
Cilium is a networking, security, and observability platform for containerized environments that leverages kernel-level data paths to process traffic. By executing programs directly within the Linux kernel, it provides high-performance packet filtering, routing, and load balancing without the need for traditional user-space proxies or context switching. The platform distinguishes itself through identity-based security enforcement, which filters traffic based on service labels rather than volatile IP addresses. It integrates containerized workloads with external physical or virtual infrastruct
Cilium leverages eBPF and kernel-level data paths for high-performance packet processing and networking in containerized environments, though it is primarily designed as a Kubernetes CNI and security platform rather than a general-purpose OS kernel acceleration library.
Open vSwitch is a software-defined network switch and high-performance data plane engine. It functions as a virtual layer 2 switch and network orchestrator, implementing the OpenFlow protocol to decouple the control plane from the data plane for programmable network forwarding. The project distinguishes itself through a userspace-kernel datapath split and the use of poll-mode drivers for kernel-bypass acceleration. It provides a network tunneling gateway to extend virtual networks across physical hosts and utilizes a transactional, schema-based configuration database with state replication fo
Open vSwitch provides a high-performance data plane engine and kernel-bypass acceleration through poll-mode drivers, making it a strong tool for packet processing even though it lacks explicit eBPF or DPDK integration in this scope.
Dpdk-ans is a high-performance userspace networking library and TCP/IP stack built on the Data Plane Development Kit. It implements a userspace network protocol stack that bypasses operating system kernel overhead and interacts directly with network interface hardware using poll mode drivers without interrupts. The project provides BSD-compatible socket APIs to enable legacy network applications to operate without requiring source code modifications. It supports multi-core receive scaling and traffic distribution to assign incoming network flows across dedicated CPU cores using hashing, alon
This repository provides a high-performance TCP/IP stack built natively on DPDK for fast packet processing, though it focuses on user-space kernel bypass rather than traditional Linux kernel modules.