lvs

LVS, or Linux Virtual Server, is an open‑source project that provides transport‑level load balancing and high‑availability services for Linux‑based networks. Developed by the China University of Technology and released under the GNU GPL, LVS implements a cluster of real servers behind a single virtual IP address, distributing client requests among them to improve performance, reliability, and scalability.

The core of LVS is the IPVS (IP Virtual Server) kernel module, which intercepts incoming packets on the load‑balancing node (the director) and forwards them to selected real servers according to a scheduling algorithm. Common algorithms include round‑robin, weighted round‑robin, least‑connections, and locality‑based least‑connections. The director can operate in three modes: NAT (network address translation), where the director rewrites packet headers; DR (direct routing), where packets are sent directly to real servers on the same subnet; and TUN (tunneling), which encapsulates packets for delivery over a separate network.

Real servers run the application services unchanged, while the virtual service is presented to clients via the virtual IP and port. Health‑checking mechanisms monitor server status, automatically removing failed nodes from the pool. LVS can be combined with tools such as Keepalived or Heartbeat to provide failover for the director itself, ensuring continuous availability.

Typical deployments include web farms, database clusters, and DNS services, where high request rates and minimal latency are critical. LVS is valued for its low overhead, kernel‑level efficiency, and flexibility, though it requires careful network design and does not perform application‑layer inspection, limiting its suitability for protocols that need deep packet analysis. The project remains actively maintained and is integrated into many Linux distributions as part of the ipvsadm and ipvsadm‑utils packages.