Directed Acyclic Graph (DAG)

What Is Directed Acyclic Graph (DAG)?

A directed acyclic graph (DAG) is a graphical representation of a sequence of events. The order of the activities is represented by a graph, which is visually portrayed as a series of circles, each of which represents an activity and is connected by lines, which indicates the flow from one activity to the next. 

Each circle is referred to as a "vertex," and each line is referred to as an "edge." "Directed" denotes that each edge has a distinct direction, implying that each edge must represent a unidirectional flow from one vertex to the next. 

The term "acyclic" refers to the absence of loops (i.e., "cycles") in the graph, which means that for any given vertex, there is no path in the network back to that initial vertex if you follow an edge connecting that vertex to another.

Uses of Directed Acyclic Graphs (DAG)

DAGs can be used to represent a wide range of flows, such as data processing operations. Considering large-scale processing processes in terms of DAGs allows one to better organize the numerous steps and associated orders for these jobs. 

A sequence of computations is executed on the data in many data processing settings to prepare it for one or more main destinations. A data pipeline is a common term for this type of data processing flow. For instance, sales transaction data might be evaluated promptly in order to make real-time recommendations to customers. 

The data can go through multiple processes during the processing lifetime, including cleansing (correcting incorrect/invalid data), aggregate (calculating summaries), enrichment (finding relationships with other relevant data), and transformation (Translating the data into a new format). 

One distinguishing feature of DAGs and the data processing processes they depict is the possibility of numerous pathways in the flow. This is significant because it recognizes the necessity to process data in various ways to meet multiple outputs and demands. For example, a sensor data stream is processed in the example flow below. 

The data is loaded from the sensors first, then split by sensor type. Data from Sensor X will be aggregated each second and examined in real-time. An alert is sent if any critical state is detected. The information is also preserved for long-term preservation and for future analysis.