Skip to main content

Azure Spot VM Billing, Pricing, and Eviction Behaviour

By

 When adopting Azure Spot Virtual Machines, one of the first challenges is understanding how billing and eviction actually work. Many people initially believe Spot VMs require a minimum runtime, such as 5 minutes. This is not correct. Spot VMs follow the same billing model as standard Pay-As-You-Go VMs: they are billed per second, with a 60-second minimum charge. If a VM runs for less than a minute, you pay for the full minute. After that, you pay only for the exact seconds it runs. The 5-minute intervals you may see on Azure pricing pages are only part of the UI graphs—they are not related to billing.

Spot pricing is based on Azure’s unused capacity. Because of this, Spot VMs often come at massive discounts—sometimes 65% to 90% off On-Demand pricing. The exact discount varies by VM family, region, availability zone, and overall demand. When provisioning a Spot VM, you can set a max price bid. Setting it to “-1” means you agree to pay up to the normal On-Demand price. Your VM will run as long as the current Spot price is equal to or below your bid. If the Spot price moves above your maximum, Azure evicts the VM. Capacity pressure can also trigger eviction, because Azure will always prioritize paying, On-Demand customers.


A critical point is that Spot VMs come with no SLA. Eviction may happen at any time. Standalone Spot VMs usually receive about 30 seconds' notice before eviction. In Virtual Machine Scale Sets, you may wait up to 2 minutes, depending on the configuration. Because of this, Spot is suited only for workloads that tolerate interruptions, such as batch processing, CI/CD agents, media rendering, GenAI model fine-tuning, big data ETL, or dev/test environments. These types of workloads must support checkpointing or quick restarts.
For workloads that must finish, it is best to use Spot as part of a resiliency strategy. A common design is to use VM Scale Sets with Spot Priority Mix mode, where Spot instances run first to save on costs and, if they are evicted, the system automatically fails over to On-Demand VMs. Another option is Azure Batch or Azure Container Instances, which can dynamically choose between Spot and On-Demand based on capacity.
Monitoring Spot's behavior is also important. You can track prices with Azure Monitor, the Pricing Calculator, or automation scripts. In regions with low contention, eviction rates may be extremely small—sometimes below 1%—which brings very high savings with low risk. Combining Spot with Capacity Reservations, Cost Management forecasting, and a solid retry/checkpoint strategy allows you to maximize both reliability and cost efficiency.
Azure Spot VMs are a strong tool for reducing cloud costs, but they require planning. When used correctly, they can deliver exceptional savings while still supporting demanding workloads.

Comments

Post a Comment

Popular posts from this blog

How to audit an Azure Cosmos DB

By
In this post, we will talk about how we can audit an Azure Cosmos DB database. Before jumping into the problem let us define the business requirement: As an Administrator I want to be able to audit all changes that were done to specific collection inside my Azure Cosmos DB. The requirement is simple, but can be a little tricky to implement fully. First of all when you are using Azure Cosmos DB or any other storage solution there are 99% odds that you’ll have more than one system that writes data to it. This means that you have or not have control on the systems that are doing any create/update/delete operations. Solution 1: Diagnostic Logs Cosmos DB allows us activate diagnostics logs and stream the output a storage account for achieving to other systems like Event Hub or Log Analytics. This would allow us to have information related to who, when, what, response code and how the access operation to our Cosmos DB was done. Beside this there is a field that specifies what was th...
Post a Comment
Read more

Can User Technical Profile Influence AI Architecture Decisions?

By
In the last few days, I ran a small experiment on Devin/Windsurf/Cascade, BMAD, and LLM-based architecture decisions. The initial concern was simple: if an AI coding assistant knows something about the user, for example, that the user is a Java developer or a .NET developer, can this information influence the architecture and technology stack that the AI will propose? This question is important because many companies are starting to use AI tools not only for coding but also for product discovery, architecture, documentation, and technical decision-making. If the AI is influenced by the developer's personal profile, the result may not be fully neutral. It may look like a business or architecture decision, but in reality, it may be partially shaped by the context of the person using the tool. The test was done using BMAD with Devin/Windsurf/Cascade. The scenario was a unified field service operations platform. The task was to let BMAD refine the product, define the architecture and s...
Post a Comment
Read more

ADO.NET provider with invariant name 'System.Data.SqlClient' could not be loaded

By
Today blog post will be started with the following error when running DB tests on the CI machine: threw exception: System.InvalidOperationException: The Entity Framework provider type 'System.Data.Entity.SqlServer.SqlProviderServices, EntityFramework.SqlServer' registered in the application config file for the ADO.NET provider with invariant name 'System.Data.SqlClient' could not be loaded. Make sure that the assembly-qualified name is used and that the assembly is available to the running application. See http://go.microsoft.com/fwlink/?LinkId=260882 for more information. at System.Data.Entity.Infrastructure.DependencyResolution.ProviderServicesFactory.GetInstance(String providerTypeName, String providerInvariantName) This error happened only on the Continuous Integration machine. On the devs machines, everything has fine. The classic problem – on my machine it’s working. The CI has the following configuration: TeamCity .NET 4.51 EF 6.0.2 VS2013 It see...
8 comments
Read more