PowerConnect Sizing Guide
PowerConnect Sizing Guide
This guide provides sizing requirements for the PowerConnect Service, helping administrators and architects understand how data volume transferred between IBM Planning Analytics (TM1) and Microsoft Power BI impacts memory consumption in the PowerConnect Service.
Use this guide to estimate hardware requirements before deployment or when scaling an existing installation.
Overview
Memory consumption in the PowerConnect Service is influenced by three key dimensions of the data retrieved from IBM Planning Analytics views:
Number of rows in the result set
Number of numeric columns (measures)
Amount of string data (e.g., element names, attributes)
All tests below were conducted by varying one parameter at a time to isolate each variable's impact on memory.
Memory Scaling: Rows
What Was Tested
Understanding how memory scales as more rows are added to a result set is critical, since row counts often represent the largest driver of memory use. Both numeric values (cell data) and string data (element IDs, attributes) contribute to the memory footprint.
Test Parameters
Parameter | Value |
|---|---|
Variable | Number of rows: 0.5M to 7M (in 0.5M increments) |
Row axis | One dimension; each element name is 10 characters long |
Column axis | One numeric measure |
Cell values | Each cell contains a value of 1 |
Results
Memory consumption scales approximately linearly as more rows are added.
The memory footprint per row decreases as total row count increases, levelling off at a lower threshold.
Average memory per row: ~320 bytes
Sizing rule of thumb: Multiply expected row count by 320 bytes to estimate memory usage for numeric row data.
Memory Scaling: Columns (Numeric Measures)
What Was Tested
This test isolates the memory impact of adding more numeric measures on the column axis of a view.
Test Parameters
Parameter | Value |
|---|---|
Variable | Number of numeric measures: 1–20 (in increments of 1) |
Number of rows | 0.5M |
Row axis | One dimension; each element name is 10 characters long |
Cell values | Each cell contains a value of 1 |
Results
Memory consumption scales approximately linearly as more numeric columns are added.
The memory footprint per column decreases as the total number of columns increases.
Average memory per column: ~97 MB
Sizing rule of thumb: Multiply expected number of measures by 97 MB to estimate memory usage for column data, at 0.5M rows.
Memory Scaling: String Data
What Was Tested
This test measures the memory impact of retrieving string data from a view — equivalent to stacking more dimensions on the row axis or using string-typed measures. String data includes element names, attribute values, and any non-numeric cell content.
Test Parameters
Parameter | Value |
|---|---|
Variable | String length per cell: 10–200 characters (in increments of 10) |
Number of rows | 0.5M |
Row axis | One dimension; each element name is 10 characters long |
Column axis | One string measure |
Results
Memory consumption increases with string length, but is not strictly linear — it reflects the total volume of characters being held in memory.
Average memory per 10 additional characters: ~9 MB (at 0.5M rows)
Sizing rule of thumb: Multiply the number of extra characters per cell (beyond 10) by 9 MB to estimate additional memory needed for string-heavy views, at 0.5M rows.
Quick Reference: Sizing Estimates
Data Type | Unit | Estimated Memory |
|---|---|---|
Rows (numeric) | Per row | ~320 bytes |
Columns (numeric measures) | Per column @ 0.5M rows | ~97 MB |
String data | Per 10 extra characters @ 0.5M rows | ~9 MB |
Key Considerations for Sizing
Combine all three factors when estimating total memory. A view with many rows, many measures, and long string values will consume significantly more memory than a view varying only one dimension.
Test with realistic data volumes from your IBM Planning Analytics environment before finalising hardware specifications.
For very large result sets (multi-million rows with multiple measures), consider distributing queries across multiple PowerConnect Service instances or limiting view scope.
Element names and attributes that are string-typed contribute meaningfully to memory — not just numeric cell data.