Identifying performance issues

Transform and Analyze Data with Microsoft Fabric

Luis Silva

Solution Architect - Data & AI

Fabric capacities

 

Diagram showing Fabric being underpinned by the OneLake storage and the compute capacity, highlighting the compute capacity

Transform and Analyze Data with Microsoft Fabric

Fabric capacities

  Diagram showing fabric items being underpinned by the OneLake storage and the compute capacity, highlighting the compute capacity being increased

Transform and Analyze Data with Microsoft Fabric

Fabric SKUs

  • Fabric SKUs are named from F2 to F2048.
  • The higher the number, the more computing power.
  • Can be compared to Power BI v-cores as a reference.

Table listing the various Fabric Azure SKUs available, from the smallest F2 to the largest F2048

Transform and Analyze Data with Microsoft Fabric

Resource sizing challenges

Diagram showing the compute resources of a Fabric capacity being depleted

  • Underestimated resources
  • Depleting computing resources leads to poor performance
Transform and Analyze Data with Microsoft Fabric

Resource sizing challenges

Diagram showing the compute resources of a Fabric capacity being depleted

  • Overestimated resources
  • Unused computing resources leads to higher costs
Transform and Analyze Data with Microsoft Fabric

Automatic resource management

  • Bursting

Bar Chart illustrating the concept of bursting: a workload spike that temporarily goes over the limit established for the capacity

Transform and Analyze Data with Microsoft Fabric

Automatic resource management

  • Bursting
  • Smoothing

Bar Chart illustrating the concept of smoothing: a workload spike is spread over a longer period of time so it doesn't exceed the capacity limit

Transform and Analyze Data with Microsoft Fabric

Automatic resource management

  • Bursting
  • Smoothing
  • Throttling

Bar Chart illustrating the concept of throttling: multiple requests that would exceed the capacity limit are rejected

Transform and Analyze Data with Microsoft Fabric

Monitoring capacity usage

Screenshot of the download page for Microsoft Fabric Capacity Metrics app

Transform and Analyze Data with Microsoft Fabric

Fabric Capacity Metrics app

  • Compute usage over time (CU%)
  • Processing time
  • Number of operations/hour
  • Number of users/hour

Screenshot of a part of the Fabric Capacity Metrics app showing the chart for CU usage, duration, number of operations and number of users

Transform and Analyze Data with Microsoft Fabric

Monitoring capacity usage

  • Compute usage over time (CU%)
  • Processing time
  • Number of operations/hour
  • Number of users/hour
  • Throttling

Screenshot of a part of the Fabric Capacity Metrics app showing the Throttling chart

Transform and Analyze Data with Microsoft Fabric

Monitoring capacity usage

  • Compute usage over time (CU%)
  • Processing time
  • Number of operations/hour
  • Number of users/hour
  • Throttling
  • Metrics per item

Screenshot of a part of the Fabric Capacity Metrics app showing the matrix by item and operation

Transform and Analyze Data with Microsoft Fabric

Monitoring capacity usage

  • Compute usage over time (CU%)
  • Processing time
  • Number of operations/hour
  • Number of users/hour
  • Throttling
  • Metrics per item
  • Storage

Screenshot of a part of the Fabric Capacity Metrics app showing the storage usage

Transform and Analyze Data with Microsoft Fabric

Monitoring hub

  • More detailed view than Capacity Metrics app
    • Capacity Metrics app: Focus on aggregated resource consumption over time
    • Monitoring hub: Focus on execution time of individual activities

 

Icons representing pipelines, dataflows, lakehouses, notebooks, semantic models and spark jobs

Transform and Analyze Data with Microsoft Fabric

Monitoring hub

Screenshot of the Monitoring hub showing a list of notebooks, dataflows and other Fabric items

Transform and Analyze Data with Microsoft Fabric

Let's practice!

Transform and Analyze Data with Microsoft Fabric

Preparing Video For Download...