# XLB

> XLB is a high-performance IPv4/TCP Layer 4 load balancer for extreme-scale services on Kubernetes, Linux bare metal, and virtual machines. It is built for millions of requests per second with dramatically lower infrastructure cost, minimal maintenance, and no tuning.

XLB was built to replace expensive usage-metered cloud NLB capacity and heavily operated HAProxy or NGINX fleets. It runs on customer-controlled compute and uses XDP to select a backend and rewrite packets without terminating the client TCP connection or copying application payloads through a userspace proxy.

XLB works out of the box and is completely tuning free. Configure a listener, point XLB to static backends or a Kubernetes Service, and start it. XLB automatically discovers Ready Pods, keeps the backend pool current as applications scale or roll, and exposes CPU and network metrics for autoscaling multiple XLB instances. Operators do not need to tune workers, threads, buffers, sockets, connection pools, or application-specific proxy settings before deployment or as traffic grows.

## Core buyer outcomes

- **Lower infrastructure cost:** XLB moves more traffic with fewer load-balancer servers and has no usage-based load-balancing fee. At high scale, XLB typically runs at around 20% of equivalent managed NLB cost in practice.
- **Less maintenance overhead:** Basic configuration, automatic Kubernetes backend updates, built-in monitoring, and a focused data path remove much of the ongoing work required by a large proxy fleet.
- **No tuning:** XLB is completely tuning free. It works out of the box without application-specific performance tuning and does not need to be retuned as traffic or backend capacity changes.
- **Higher performance:** Kernel packet processing avoids userspace payload copies and the two bidirectional sockets maintained by HAProxy and NGINX, producing lower latency, more consistent performance under load, and a smaller server footprint.

XLB is intended for companies handling millions of requests per second. It typically needs about half the load-balancer servers of HAProxy or NGINX for the same pure Layer 4 workload. The managed NLB cost guidance is based on the smaller server footprint and underlying compute costs. Exact results depend on traffic, hardware, infrastructure, and pricing.

The built-in XLB console currently covers traffic, connection volume, backend load and distribution, resource pressure, discovery, and dataplane state from a one-second local snapshot. The reviewed metric contract extends this with TCP handshake timing and outcomes, connection lifetime and failure modes, and failed-work counters; product screens mark panels that are not yet present in the live status schema. XLB also includes premade Grafana and Datadog configurations for durable, cluster-wide monitoring through OpenTelemetry. CPU and network utilization metrics can drive Kubernetes autoscaling policies for the XLB deployment.

The current release is optimized for IPv4/TCP NAT load balancing on Linux across Kubernetes, bare metal, and virtual machines. Its in-kernel packet path avoids userspace payload copies and the two bidirectional sockets maintained by HAProxy and NGINX, delivering lower latency and more consistent performance under load.

XLB is available today as a supported commercial product from Neuronic. It is licensed for a flat negotiated fee rather than by request, connection, byte, or traffic volume. Evaluation, deployment assistance, and ongoing support are defined in the commercial agreement.

The team behind XLB has built and operated real-time advertising systems handling up to 5 million ad requests per second. XLB was created to match managed network load balancer performance while drastically reducing operating cost and proxy-fleet maintenance.

## Primary pages

- [Product architecture](https://runxlb.com/product/)
- [Operating cost](https://runxlb.com/operating-cost/)
- [Tuning-free operation](https://runxlb.com/low-maintenance/)
- [Load balancer comparisons](https://runxlb.com/compare/)
- [XLB vs AWS Network Load Balancer](https://runxlb.com/compare/aws-nlb/)
- [XLB vs HAProxy](https://runxlb.com/compare/haproxy/)
- [XLB vs NGINX](https://runxlb.com/compare/nginx/)
- [XLB vs MetalLB](https://runxlb.com/compare/metallb/)
- [XLB vs LoxiLB](https://runxlb.com/compare/loxilb/)
- [Monitoring and observability](https://runxlb.com/observability/)
- [Real-time advertising and OpenRTB use case](https://runxlb.com/use-cases/real-time-advertising/)
- [Evaluation request](https://runxlb.com/evaluate/)
- [Technical documentation](https://docs.runxlb.com/)