This chapter covers Azure Spot Virtual Machines, a cost-saving option for workloads tolerant of interruption. AZ-900 tests this under objective 2.2, as a cost-related compute option.
Jump to a section
A simple way to picture Azure Spot Virtual Machines
Airlines sell seats at regular price and also sometimes offer deeply discounted seats that the airline can cancel (to sell at full price) if demand spikes. Azure Spot Instances work similarly — they're heavily discounted virtual machine pricing, in exchange for the possibility that Azure can reclaim them if capacity is needed elsewhere. You save money by betting that you can tolerate the interruption.
What Spot Instances are
Azure Spot Virtual Machines are virtual machines available at a significantly discounted price (often 70-90% off standard pricing) in exchange for the acceptance that Azure can reclaim them at any time if capacity is needed for standard, full-price instances.
Suitable and unsuitable workloads
Spot Instances fit workloads that can tolerate interruption without significant impact — batch processing jobs, non-critical development/test environments, or analysis workloads. They don't fit production services or workloads requiring guaranteed availability, since interruption would be problematic.
How they fit cost management strategy
For workloads that can use them, Spot Instances dramatically reduce cloud costs, fitting the broader consumption-based pricing and cost optimization themes covered in earlier chapters.
They're not guaranteed to be available
Availability depends on capacity and demand — sometimes Spot capacity is unavailable at all, and you'd need to use standard instances instead.
Identify suitable workloads
Workloads that can tolerate interruption are candidates for Spot Instances — batch jobs, development/test, analysis.
Accept the tradeoff
Accept that the instance can be reclaimed at any time in exchange for the steep discount.
Handle interruption gracefully
Workloads should be designed to handle interruption gracefully — save state frequently if data loss on interruption would be problematic.
Set an eviction policy
Configure how the instance should behave when reclaimed — either deallocate (pause and try to restart later) or delete (terminate completely).
A company running batch analysis jobs on a schedule uses Spot Instances for most of the computation, saving 80% on compute costs compared to standard instances. Jobs that fail due to interruption are automatically restarted, tolerating the occasional interruption.
Objective 2.2 expects candidates to understand Spot Instances as a cost-saving option for workloads tolerant of interruption, not a universally applicable choice for every workload.
A common wrong answer is assuming Spot Instances are always cheaper without recognizing the interruption tradeoff — they're only appropriate for specific workload types.
Stable terms: Spot Instances, interruption. Memory trick: Spot Instances = huge discount (70-90% off), but Azure can reclaim them anytime — only suitable for workloads tolerant of interruption.
Azure Spot Instances offer steep discounts (70-90% off standard pricing) in exchange for accepting that Azure can reclaim them anytime.
They're suitable only for workloads tolerant of interruption — batch processing, development/test, analysis.
They're not appropriate for production services or workloads requiring guaranteed availability.
Spot capacity is available only when Azure has surplus capacity — availability is not guaranteed.
These come up on the exam all the time. Here's how to tell them apart.
Standard Virtual Machines
Full price
Guaranteed availability
Suitable for all workloads
Right choice for production/availability-critical
Spot Instances
70-90% discount
Can be reclaimed anytime
Suitable only for interruption-tolerant workloads
Right choice for cost-sensitive batch/dev/test
Mistake
Spot Instances are the right choice for all cost-sensitive workloads.
Correct
They're only appropriate for workloads that can tolerate interruption — production services or availability-critical workloads shouldn't use them.
Mistake
Spot Instance availability is guaranteed.
Correct
Spot capacity is available only when Azure has surplus capacity — sometimes it's unavailable, and you'd need to use standard instances instead.
They're virtual machines available at steep discounts (typically 70-90% off standard pricing) in exchange for accepting that Azure can reclaim them at any time if capacity is needed elsewhere.
Workloads that can tolerate interruption without significant impact — batch processing jobs, development/test environments, analysis workloads. Production services or availability-critical workloads should not use them.
No — Spot capacity depends on Azure's available surplus capacity. Sometimes Spot is unavailable, and you'd need to use standard instances instead.
You've just covered Azure Spot Virtual Machines — now see how well it sticks with free AZ-900 practice questions. Full explanations included, no account needed.
Done with this chapter?