How Load Balancers Route Web Traffic Based on Specific Content

A method for web servers to route user requests to specific machines based on which files they store, rather than just blindly balancing traffic across all servers.

What it covers

This patent describes a system where a central load balancer acts as a gatekeeper for a website. Instead of mirroring every file on every server, the system distributes specific resources across different nodes. When a user requests a file, the load balancer intercepts the request, reads the URL to see exactly what is being asked for, and then routes the connection to a server that actually holds that specific file. Once the connection is handed off, the chosen server sends the data directly back to the user, bypassing the load balancer to prevent network bottlenecks.

What it doesn't cover

• —Does not cover systems where every server contains an identical copy of all website resources (full mirroring).
• —Does not cover load balancing methods that assign servers based solely on CPU load or round-robin traffic distribution without inspecting the requested URL.
• —Does not cover systems where the load balancer remains in the data path for the entire duration of the file transfer.

The clever bit

The system delays the load balancing decision until after the URL is parsed, allowing the network to make an intelligent routing choice based on content location rather than just network capacity.

Why it matters

This technology was essential for the early scaling of the World Wide Web. By allowing web hosts to partition content across cheaper, specialized servers rather than requiring massive, fully-mirrored server clusters, it made hosting large, complex websites economically viable during the late 1990s dot-com boom.

Real-world examples

• 1.Modern content delivery networks (CDNs)
• 2.Distributed web server clusters
• 3.Enterprise load balancing appliances