PowerConnect Sizing Guide

This guide aims to assist with sizing requirements for the PowerConnect Service. Below you will find information about how data volume transferred between IBM Planning Analytics and Microsoft Power BI impacts memory consumption in the PowerConnect Service.

Tests were performed to take some of the variability out of the picture, meaning that only one parameter (e.g., number of rows and columns) was changed during each test.

Even though the numeric values you transfer from IBM Planning Analytics to Microsoft Power BI are critical to providing end-users insight into organizational data, element information, such as element IDs and attributes, is equally important from a sizing perspective. When it comes to understanding the memory footprint of the PowerConnect Service, it is relevant to see the difference between what impacts transferred numeric vs. string data (e.g., element IDs).

Testing Scenario

This memory scaling test was performed with the following parameters:

• Variable: Number of rows: 0.5M to 7M in 0.5M increments.
• One dimension on the row axis, each element being 10 characters long.
• One numeric measure on the column axis.
• Each cell contains a value of 1.

The chart above shows a fairly linear progression of memory consumption when adding more rows.

The chart above shows the number of bytes per row. A larger number of rows will result in a smaller memory footprint per row until the lowest threshold is reached.

The average number of bytes per row in this test is 320.

This test was performed to understand how memory scales when adding more numeric columns to a view.

Testing Scenario

This memory scaling test was performed with the following parameters:

• Variable: Number of numeric measures on the column axis: 1-20, incremented by 1.
• Number of rows: 0.5M.
• One dimension on the row axis, each element being 10 characters long.
• Each cell contains a value of 1.

The chart above shows a fairly linear progression of memory consumption when adding more columns (measures).

The chart above shows the number of bytes per columns. A larger number of numeric measures (columns) will result in a smaller memory footprint per column until the lowest threshold is reached.

The average number of megabytes per column in this test is 97.

This test was performed to understand how memory scales when adding more characters are retrieved in a view. This would be equivalent to adding more dimensions stacked on the row axis in a view.

Testing Scenario

This memory scaling test was performed with the following parameters:

• Variable: Each cell contains string data between 10 and 200 characters in increments of 10.
• Number of rows: 0.5M.
• One dimension on the row axis, each element being 10 characters long.
• One string measure on the column axis.

The chart above shows the memory consumption when adding more string data in a view.

The average number of megabytes per each added 10 characters in this test is 9.

PowerConnect is available on a 30-day free trial with limited release to North American customers only at this time. Contact us to learn more.