XLB / the world’s fastest Layer 4 load balancer
XLB is a high-performance Layer 4 load balancer for Kubernetes, bare metal, and VMs. Built for companies handling millions of requests per second, it cuts operating cost and maintenance overhead with fewer servers, no traffic-based fees, and no proxy tuning.
Why XLB / the business case
Managed NLB pricing grows with traffic. HAProxy and NGINX replace that usage fee with more servers, ongoing operational work, and potentially commercial licensing or support. XLB reduces the total cost.
XLB moves more traffic per server and runs on infrastructure you control. At high scale, XLB typically runs at around 20% of equivalent managed NLB cost in practice and needs about half the load-balancer servers of HAProxy or NGINX for the same pure L4 workload.
Configure a listener and point XLB at a backend list or Kubernetes Service. XLB is completely tuning free, with no separate VIP, BGP, failover, or configuration control plane to maintain.
Why XLB exists
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 infrastructure cost and the maintenance burden of large proxy fleets.
Neuronic continues to build high-performance advertising infrastructure. XLB is the load balancer we needed for that work, packaged as a supported commercial product for other teams operating at the same scale.
About NeuronicDeploy once / Kubernetes or bare metal
Give XLB a Kubernetes namespace and Service name. It finds the selected Pods, sends traffic only to ready backends, and keeps the pool current as your deployment scales and rolls. On bare metal, give it a short list of backend IPs instead.
Configure it once. No tuning. No separate control plane.
listen: auto
ports:
- local_port: 443
remote_port: 8443
provider:
kubernetes:
namespace: production
service: bidder-apiArchitecture
HAProxy and NGINX terminate the client connection, pass application data through a proxy process, and establish a second connection to the backend. XLB keeps traffic on the kernel forwarding path and routes packets without copying payloads through userspace.
Less data movement and connection management means lower latency, more consistent performance under load, and higher throughput per server.
Explore the architectureLess data movement is why XLB can deliver higher throughput with fewer servers.
Use case / real-time advertising
Large ad platforms handle huge request volume under strict timeouts. AWS NLB turns that volume into a large monthly bill. HAProxy and NGINX avoid that usage fee, but still require more servers, tuning, monitoring, and potentially commercial licensing or support. XLB significantly reduces both the operating cost and maintenance burden of the load-balancing layer.
Real questions
Yes. Configure a namespace and Service name. XLB watches the Service selector, adds Ready Pods automatically, and stops assigning new connections to Pods that become unready or start terminating. There is no backend list to maintain as your application scales or rolls.
Yes. Run multiple XLB replicas across Kubernetes nodes for capacity and redundancy. XLB exposes CPU and network utilization metrics for Kubernetes autoscaling. Each replica watches the same Service and maintains its own current backend set as capacity changes.
For planning, XLB typically needs about half the load-balancer servers of HAProxy or NGINX for the same pure L4 workload. It is designed to reach line rate without application-specific tuning. Exact throughput depends on the NIC, driver, packet size, and connection rate, so we validate the comparison on the customer’s hardware.
At high scale, XLB typically runs at around 20% of the cost of an equivalent managed NLB deployment in practice. That guidance comes from the smaller server footprint and underlying compute costs. The exact result varies with traffic, hardware, region, and pricing.
XLB is a commercial product licensed for a flat negotiated fee. Pricing is not based on requests, connections, bytes, or traffic volume, so the software bill remains predictable as traffic grows.
Yes. XLB is available today from Neuronic as a commercially licensed product. Evaluation, deployment assistance, and ongoing product support are defined as part of the commercial agreement.
Run the numbers
Bring your cloud bill or proxy topology. We’ll scope an evaluation around your traffic, hardware, and the operational work you want to eliminate.