Skip to main content

Katran

Katran is an open-source layer 4 load balancer (networking / traffic management) based on eBPF and XDP, originally developed by Meta and released through the facebookincubator organization.

  • High-performance L4 load balancing on top of XDP and eBPF (network traffic management)
  • Implements a scalable L4 load balancer in kernel space with a userspace control plane (network infrastructure)
  • Supports direct server return and connection tracking models for service traffic (application delivery)
  • Provides reusable eBPF/XDP data plane library code for custom load-balancing solutions (eBPF tooling)
  • Targets deployment on commodity servers running Linux as a software load balancer (infrastructure software)

More About Katran

Katran is an eBPF- and XDP-based layer 4 load balancer (network traffic management) designed to run on commodity Linux servers as a software load balancer. It originates from Meta and is published in the facebookincubator GitHub organization. The project focuses on enabling high-throughput, low-latency packet processing at the edge of the network by using eBPF programs attached at the XDP hook to process packets early in the kernel networking stack.

The core of Katran is a data plane implemented as eBPF programs (eBPF tooling) that perform layer 4 load balancing decisions directly in the kernel. These programs can parse incoming packets, apply consistent hashing or other load distribution algorithms, and redirect traffic to backend servers according to configured service maps. The data plane is coupled with a userspace component (network infrastructure) that manages configuration, such as backend pool membership, health status, and routing policies, by updating eBPF maps that the kernel programs consult at runtime.

Katran uses XDP (Express Data Path) (Linux networking) to run its eBPF code as early as possible in the packet reception path, which reduces overhead from the traditional kernel networking stack. This architecture allows Katran to handle high packet rates while maintaining flexibility through its control plane. The project provides libraries, example configurations, and reference code for building customized load balancing solutions, including support for features such as direct server return (DSR) and connection-aware load distribution where enabled.

In enterprise and large-scale environments (infrastructure software), Katran can be deployed as a software load balancer in front of web or Application Programming Interface (API) services, often replacing or complementing hardware appliances. Operators can integrate Katran with existing service discovery or health checking systems by driving updates to its eBPF maps from their own control-plane software. Because it is built on standard Linux kernel facilities, Katran aligns with infrastructure based on commodity x86 or similar servers and integrates with the broader eBPF ecosystem.

From a categorization standpoint, Katran fits into network traffic management and application delivery, with a focus on L4 load balancing using eBPF/XDP on Linux. Its architecture demonstrates a model where the data plane resides in eBPF in the kernel, while policy and configuration reside in userspace processes updating maps. For enterprise technical stakeholders, Katran offers a programmable, open-source load balancing data plane suitable for environments that adopt eBPF-based networking.