Find Skew in Distributed Tables
Bradley Schacht is a Principal Program Manager on the Microsoft Fabric product team based in Saint Augustine, FL. Bradley is a former consultant, trainer, and has coauthored 6 SQL Server and Power BI books, most recently Learn Microsoft Fabric. As a member of the Microsoft Fabric product team, Bradley works directly with customers to solve some of their most complex data problems and helps shape the future of Microsoft Fabric. Bradley gives back to the community through speaking at events such as the SQLBits, Fabric Community Conference, PASS Community Data Summit, SQL Saturdays, Code Camps, and user groups across the country including locally at the Jacksonville SQL Server User Group (JSSUG). He is a contributor on SQLServerCentral.com and blogs on his personal site, BradleySchacht.com.
Let’s examine the results. We end up with a record for each partition, node, and distribution combination.
Now we have the same results (in my case 4,160 rows = 52 Partitions * 10 Compute Nodes * 8 Distributions per node). Next, we will create a PivotTable and PivotChart from the table. In Excel on the Insert tab select PivotTable from the far left. Select to add this table to a new worksheet. From the PivotTable field list drag DISTRIBUTION_ID to rows followed by PDW_NODE_ID to rows. Then drag ROW_COUNT to Values. By default this will be a count of ROW_COUNT, not a sum. To switch it to a sum click the arrow on the right side of ROW_COUNT and select Value Field Settings. Change the Summarize value field by from Count to Sum. 
The result should look like this: 
Now we have a count of rows in each distribution. Ideally this should be as close to the same number in each distribution as possible. My table happens to have an awesome distribution key and the skew is less than 1,000 from the smallest to largest distribution. Not bad for a table with over 48 million records! Let’s take it one step further though. We want to not only see the numbers, but have a visual of our skew. With the PivotTable selected go to the Insert tab and select column chart. Doing this will tie the column chart to the same data that is in the PivotTable. Initially our data doesn’t look so good. There are some large peaks and valleys in the chart, but we know that the skew isn’t bad. If you look at the scale for the chart you’ll notice that Excel has “intelligently” decided to start the scale at a value that will allow us to see the difference in each bar. It does not start at 0. 
If you right-click on the Y-axis and select Format Axis you can change the minimum bound from the automatically chosen value to 0. Now the skew in the chart is a much better representation of the data in the table. 
From here you can change the filters and look at skew inside each individual node or a combination of distributions/nodes. I have posted a copy of the Excel sheet I created so you have a reference to use later on. One last note: for distribution keys try to choose a column with no NULL values or very few. NULL will always hash to Compute Node 1, Distribution 1 no matter the data type.