Cloud Networking Glossary
Learn the Fundamentals

What is Network Load Balancing (NLB)?

Network Load Balancing is an availability technique that allows a resource to be served over multiple WAN-link channels that divide the sending payload across application components or transported segments. This technique can be classified into two types of Load Balancing that are referred to as Layer 4 or Layer 7. As layer 7 refers to the application layer of the OSI model, the application itself is the division frame of reference when the NLB component is provisioned to make the determinations over which requests should be served from which node of the NLB.

You may have experienced this first hand when you selected a URL in the browser and waited a few seconds for it to load, and noticed the data components or widgets show up in the browser arranged one way, only for the page to seemingly refresh itself so that the page then appears normally. This is because sometimes the HTML and images of a website may be served asynchronously with the CSS and javascript content across two different nodes of a given load balancer. 



Now layer 4 load balancing is a little more under the hood than just dividing up the work amongst the different folders of a web asset. When we refer to layer 4 of the OSI model, we are referring to the part of the payload transmission that is broken up into elements referred to as TCP segments. The TCP ( transmission control protocol ) layer of the internet’s communication paradigm is precisely calibrated for just this kind of adventure because it was designed with two specific features called Reliability and Flow Control.

Because of these two features, the rate and throughput predictability of which segments to send at which times is highly governable. The reliability feature allows for the acknowledgment of receipt of a segment or segments before the next batch is sent, and the flow control decides how many to send at one time. With each of these features readily available, the segmentation of the payload can be chopped up - or load balanced - in an even and orderly fashion to transmit the data in parallel across multiple channels on the network at once.