Spot Instances — Technology Wiki

Overview

Direct Answer

Spot instances are unused cloud computing capacity that providers offer at discounts typically 70–90% below on-demand rates, with the trade-off that the provider can terminate the instance with minimal notice when capacity is needed elsewhere. This model allows organisations to access compute resources opportunistically rather than maintaining guaranteed availability.

How It Works

Cloud providers maintain spare infrastructure capacity across data centres. When demand for standard reserved or on-demand instances declines, they release this surplus capacity as interruptible compute at auction-like pricing. Customers specify a maximum price they will pay; if the provider's available capacity falls below that threshold, instances are reclaimed, typically with a two-minute termination warning.

Why It Matters

The cost reduction is substantial for non-time-critical workloads, enabling organisations to run larger-scale batch processing, machine learning training, and analytics without proportional budget increases. This elasticity supports innovation and testing in resource-constrained enterprises whilst allowing providers to optimise utilisation of their infrastructure.

Common Applications

Batch data processing, long-running machine learning model training, distributed rendering, genomic sequencing analysis, and retrospective log analysis benefit significantly from spot capacity. Development and testing environments also leverage spot instances to reduce infrastructure costs without affecting production systems.

Key Considerations

Interruption risk makes spot instances unsuitable for stateful, long-running services or time-sensitive workloads unless paired with automatic failover and fault-tolerance mechanisms. Practitioners must architect applications with graceful termination handling and implement checkpointing for iterative workloads.

Cross-References(1)

Cloud Computing

Related in Service Models

Cloud Computing

The delivery of computing services — servers, storage, databases, networking, software — over the internet on demand.

Infrastructure as a Service

Cloud computing model providing virtualised computing resources like servers, storage, and networking over the internet.

Platform as a Service

Cloud computing model that provides a platform for developers to build, deploy, and manage applications without managing infrastructure.

Software as a Service

Cloud computing model that delivers software applications over the internet on a subscription basis.

Function as a Service

A serverless cloud computing model where individual functions are executed in response to events.

Serverless Computing

A cloud execution model where the provider dynamically allocates resources, charging only for actual compute time used.

Cloud-Native

An approach to building applications that fully exploit cloud computing advantages like elasticity, resilience, and automation.

Private Cloud

Cloud computing resources used exclusively by a single organisation, either on-premises or hosted by a third party.

Public Cloud

Cloud computing resources shared among multiple organisations and available to the general public over the internet.

Managed Service

A cloud service where the provider handles infrastructure management, maintenance, updates, and monitoring.

Cloud Cost Optimisation

Strategies and practices for minimising cloud computing expenses while maintaining performance and functionality.

Spot Instance

A cloud computing option that uses spare capacity at significantly reduced prices with the possibility of interruption.

More in Cloud Computing

Cloud-Native Database

Strategy & Economics

A database designed from the ground up to operate in cloud environments with automatic scaling and high availability.

REST API

Architecture Patterns

An API architectural style using HTTP methods and stateless communication for web service interaction.

Kubernetes

Infrastructure

An open-source container orchestration platform for automating deployment, scaling, and management of containerised applications.

Auto-Scaling

Infrastructure

Automatically adjusting compute resources based on current demand to maintain performance and optimise costs.

Hybrid Cloud

Strategy & Economics

An IT architecture combining on-premises infrastructure with public and private cloud services.

Reserved Instance

Strategy & Economics

A cloud pricing model where users commit to a specific resource configuration for a term in exchange for discounted rates.

Cloud Bursting

Strategy & Economics

A configuration where an application runs in a private cloud and bursts into a public cloud when demand spikes.

Serverless Database

Strategy & Economics

A database service that automatically provisions, scales, and manages infrastructure on demand without manual server management.