Schedules, rightsizing, or reservations: which cost tool should I use?

Prev Next

Overview

Cost Analyzer offers several overlapping optimization tools — schedules, rightsizing, reservations, savings plans, hybrid benefits, and spot instances. They can all reduce your Azure bill, which makes them seem interchangeable. They aren't. Each tool targets a specific type of waste, and applying them in the wrong order, or to the wrong workloads, either produces minimal savings or creates new problems.

Business value

Knowing which tool fits your situation means acting on the highest-impact opportunity first — not the most accessible one.

The core logic

These tools work at different layers:

  • Schedules — eliminate idle runtime. The resource exists and is correctly sized; it just shouldn't be running right now.
  • Rightsizing — eliminate over-provisioning. The resource runs continuously but is larger than the workload needs.
  • Reservations and savings plans — lock in discounted rates for resources that will run continuously at their current size.
  • Hybrid benefits — reduce licensing costs on existing Windows Server and SQL Server workloads.
  • Spot instances — use spare Azure capacity at deeply discounted prices for interruptible workloads.

Critical sequencing rule: always stabilize before you commit. Apply schedules and rightsizing first. Purchasing a reservation for an oversized VM that will be rightsized next month wastes the commitment — you pay the discounted rate for the wrong size.

Decision guide

Schedules — use when the resource is idle on a predictable pattern

A schedule starts and stops (or scales up and down) a resource based on defined business hours. It delivers immediate, measurable savings for any resource that runs continuously but only needs to exist during working hours.

Use schedules when:

  • The resource is a development, test, or non-production environment
  • The resource runs on a predictable schedule (business hours, batch window, nightly job)
  • The workload can tolerate downtime outside those hours
  • Dependent services handle unavailability gracefully

Avoid schedules when:

  • The resource is production and downtime is not acceptable without a tested runbook
  • The workload has unpredictable activation patterns — a scheduled startup may miss demand

Supported resource types include Virtual Machines, SQL Databases, App Service Plans, Cosmos DB, Logic Apps, and 8 more.

Rightsizing — use when the resource is over-provisioned for its actual workload

Rightsizing analyzes CPU, memory, and disk utilization and recommends a smaller SKU when the resource is consistently underutilized. Basic recommendations surface Azure Advisor insights directly in Cost Analyzer. Advanced recommendations let you configure observation periods and account for peak utilization rather than just averages.

Use rightsizing when:

  • A resource consistently runs at a fraction of its provisioned capacity
  • A resource has been idle for an extended period (deletion candidate)
  • You want to reduce baseline costs without changing when the resource runs

Approach carefully when:

  • The resource has burst workloads or unpredictable usage patterns — use Advanced recommendations, which accounts for peaks
  • The resource is production — validate the recommended SKU in a non-production environment first
  • The resource is already covered by a reservation — changing the SKU can leave the commitment misaligned

Rightsizing actions are applied directly in Azure and are not automatically reversible from within Cost Analyzer. Always test in non-production first.

Reservations — use when a workload will run continuously at a stable SKU

A reservation commits to a specific resource type, SKU, and region for 1 or 3 years in exchange for a discount of up to 72% compared to pay-as-you-go pricing. Cost Analyzer surfaces reservation recommendations based on your actual usage patterns.

Use reservations when:

  • The workload runs 24/7 with predictable, stable SKU and regional requirements
  • The resource has already been rightsized (don't reserve an oversized resource)
  • You have good confidence in the workload's 1–3 year trajectory
  • The resource type supports reservations (Virtual Machines, SQL Databases, Cosmos DB, and others)

Avoid reservations when:

  • The workload is variable in SKU, region, or service type — use savings plans instead
  • The resource is scheduled to be rightsized in the near term
  • You're uncertain about the workload's lifespan

Prefer 1-year terms until utilization patterns are well-established.

Savings plans — use when compute spend is consistent but resource mix is variable

A savings plan commits to a consistent hourly compute spend (for example, $5/hour) across any eligible resource type, region, or service. It's more flexible than a reservation but offers a slightly smaller discount.

Use savings plans when:

  • Compute spend is predictable in amount but varies in which resource types or regions carry it
  • Your organization is actively migrating between VM families or Azure regions
  • Reservations are too specific for your current workload mix

Hybrid benefits — apply immediately to eligible workloads

Azure Hybrid Benefit uses existing on-premises Windows Server or SQL Server licenses to reduce the cost of running those workloads in Azure. Cost Analyzer identifies eligible resources and surfaces recommendations.

Use hybrid benefits when:

  • You have eligible Windows Server or SQL Server licenses with Software Assurance
  • Resources running those workloads are not already using Hybrid Benefit

This is a license-based discount with no usage risk. Apply it immediately to every eligible resource — there is no reason to delay.

Spot instances — use only for fault-tolerant, interruptible workloads

Spot instances use spare Azure capacity at significantly reduced prices. Azure can reclaim this capacity with short notice when demand rises.

Use spot instances when:

  • The workload is batch processing, rendering, data transformation, or a dev/test environment that can restart without data loss
  • Interruption is acceptable and the workload can handle it gracefully

Never use spot instances for:

  • Production workloads
  • Latency-sensitive applications
  • Stateful services that cannot tolerate interruption

Recommended sequencing

Apply these tools in order, not all at once:

  1. Establish visibility first. Use cost analysis to identify your top 5–10 cost drivers. Act on the highest-impact opportunities, not the most accessible ones.
  2. Apply schedules to development, test, and non-production resources that don't need to run outside business hours. Savings are immediate.
  3. Apply rightsizing progressively, starting with non-production environments. Use Advanced recommendations for workloads with variable utilization.
  4. Apply hybrid benefits to all eligible workloads immediately. No risk, no delay.
  5. Purchase reservations and savings plans for workloads that have been stabilized — after schedules and rightsizing are in place.
  6. Evaluate spot instances only for workloads you've confirmed are genuinely fault-tolerant.

Quick reference

I want to reduce spend on... Tool to use
Dev/test resources running overnight and weekends Schedules
VMs or databases that are consistently underutilized Rightsizing
Resources that run 24/7 at a predictable SKU Reservations
Consistent compute spend across changing resource types Savings plans
Windows Server or SQL Server workloads Hybrid benefits
Batch processing or rendering jobs Spot instances

Related articles