Making data-driven decisions

Data analysis became the new gold rush not long ago, and the analysis outcomes became the gold itself. At the same time, businesses started to do data-driven decision-making (DDDM) by making decisions based on tangible results from solid data analysis instead of guess-estimating for business directions. SREs employ a similar approach to choosing from disparate technical courses of action, one of which will yield a better system overall and increase reliability.

SREs need to answer many questions daily, for instance, about a system’s performance degrading over time. We feel tempted to probe the system a few times and come up with quick conclusions. Although this form is a fast way to answer questions, it lacks precision. This usually leads to empty efforts and unnecessary expenses.

There are trends and patterns in the data that span out over a daily timeframe. Even though SREs stay watching graphs for a whole day, they will not detect a shape...