Conceptual Underpinnings of Load Balancing

The primary goal of load balancing is to improve throughput. Load balancing requires multiple routes or links, which often are implemented to improve redundancy. Multiple levels of load balancing may be implemented simultaneously. Indeed, this is often the case. Routing protocols often load-balance across multiple links, some of which are logical links composed of multiple physical links. Likewise, session-oriented protocols often load-balance across multiple sessions, while each session is composed of multiple Transport Layer connections. To avoid undesirable behavior in such environments, network and storage administrators must understand the operation of each protocol that performs load balancing and the interaction between those protocols.

The phrase load sharing is often used interchangeably with the phrase load balancing, but some people distinguish between these phrases. Load sharing can mean that multiple paths are used with each path bearing an arbitrary load. By contrast, load balancing can imply that some effort is made to balance the traffic load evenly across each available path. This distinction is important to note when reading white papers, design guides, and the like. So, readers are encouraged to consider the context when either of these terms is used. Both terms are used interchangeably within this book.