97 Things Every SRE Should Know

Chapter 60. The Importance of Margins in Systems

Kurt Andersen

LinkedIn

Margin is a tool for handling uncertainty and one of the multiple objectives that needs to be balanced when managing a technical or human system. In fields ranging from psychology to mechanical engineering to queueing theory and reliability engineering, the ability of a system to adapt (and in some cases, even perform at all) is critically dependent on the time and space buffers that are part of the system. Just as I’ve discussed the value of margins for individuals, margins are important for systems too.

Margins must be included in the design of functional systems to account for uncertainty and the additive effect of individual system tolerances. Computer networking is a great example of where the margin calculations are a standard part of the practice. Ensuring that every network link in the path has excess capacity beyond the expected bandwidth usage is a critical part of managing network performance. The older rule of thumb was to upgrade any links that were averaging 50% utilization; part of that was to account for lead times in the procurement process, but part of it was to handle unpredictable peaks in instantaneous traffic loads gracefully.

Newer guidelines, along with measures such as QOS (quality of service) prioritization, have allowed network engineers to push the average utiilization into the ...