The Data Engineer's Guide to Lightning-Fast Apache Superset Dashboards

One of the most common complaints people have after using a business intelligence (BI) tool for a while is that “dashboards are slow.” No matter how beautiful and well constructed a dashboard is, or how well presented the information is, if the charts take minutes to load the dashboard becomes useless. In fact, according to research conducted by IBM in the '60s, human users become impatient if they have to wait more than π (3.14) seconds for a response from an interactive computer system — a unit jokingly described as a nanocentury.

How can we achieve those speeds? Is it really possible to make dashboards load in less than four seconds when we’re looking at metrics computed from millions or billions of rows? To understand how we can obtain those results, first let’s take a look at the lifecycle of a dashboard, starting from the initial SQL queries that power charts.

Data visualization usually starts with an exploratory phase, where the user is still unsure of what information they want to present. Imagine a table orders with records of all the products sold by a given company. A data analyst might start by looking at the overall volume of sales by running the following query:

Leave a Comment