High Azure VM Usage Reporting due to Instance Size Flexibility (ISF)
search cancel

High Azure VM Usage Reporting due to Instance Size Flexibility (ISF)

book

Article ID: 446315

calendar_today

Updated On:

Products

CloudHealth

Issue/Introduction

Customers may observe an unexpected spike or jump in Virtual Machine (VM) usage metrics for a specific month.  While the actual costs remain unaffected or consistent with previous periods.

Environment

  • Product: CloudHealth by Broadcom
  • Cloud Provider: Microsoft Azure
  • Feature: Virtual Machine Reserved Instances (RIs)

Cause

This behavior is often tied to the Instance Size Flexibility (ISF) setting on Azure Reservations.

When ISF is enabled, the reservation discount is shared across a group of different VM sizes rather than being locked to a single specific instance size. In billing and usage reporting:

  • Azure generates a separate line item for every VM that receives a portion of that reservation discount.
  • If multiple VMs utilize the discount, or if a single VM utilizes it in different ways, the cost report creates multiple entries.
  • This can lead to what appears to be "duplicate" or inflated usage entries in reporting tools like CloudHealth, even though the net cost impact is zero (or consistent with the reservation price).

Resolution

In CloudHealth you can run the following report, to showcase Instance Size Flexibility (ISF) is active on an Azure Reservation and how Microsoft logs these Azure log items:

  • Virtual Machine RI Utilization
    • To see this play out clearly across your environment without the complexity of writing custom FlexReports, we have configured a specific view on your standard out-of-the-box Virtual Machine RI Utilization report.

      By applying the filter to include only "On Demand" Reservation Types, the visual trend on the chart tells the entire story:

      • The Spiked Blue Bars (Compute Cost Before RI): This tracks what your infrastructure would have cost at standard On-Demand rates.  This confirms that Azure's billing engine was processing a huge wave of dynamic, split line items as it calculated your flexibility ratios.

      • The Flat Red Bars (Compute Cost): This tracks what you actually paid after the reservation discount was applied. Notice how the red bars stay completely steady and flat during January 2026, completely ignoring that massive blue spike.

      This layout visually proves the scenario: Your virtual infrastructure processed more reporting line items (the Blue spike), but your final bill didn't budge because your reservation successfully absorbed those lines down to a net cost of £0.00 (the flat Red baseline).