What Are Spot Instances? Affordable Compute for Indian Workloads

Spot instances are unused cloud compute capacity sold at steep discounts compared to on-demand pricing. According to AWS, Spot Instances offer savings of 60-90% compared to on-demand rates. The trade-off is that the cloud provider can reclaim (interrupt) your instance with a short notice period, typically two minutes on AWS. For Indian workloads that can tolerate interruptions, spot instances are the most cost-effective compute option available.

Key Takeaways

• Spot instances provide 60-90% savings over on-demand pricing (AWS).
• Best suited for batch processing, data analytics, CI/CD pipelines, and testing environments.
• AWS Mumbai and Hyderabad regions both offer spot capacity for Indian workloads.

[INTERNAL-LINK: cloud cost optimization → cloud cost optimization]

How Do Spot Instances Work?

Cloud providers have fluctuating capacity. When demand for a particular instance type in a specific region drops, excess capacity becomes available as spot instances. According to Datadog's 2025 infrastructure report, spot instance adoption grew by 28% year-over-year across APAC. The provider sells this capacity at a discounted rate, and you use it until either you terminate the instance or the provider needs the capacity back.

AWS Spot Instances

AWS Spot Instances in the Mumbai (ap-south-1) and Hyderabad (ap-south-2) regions typically offer 65-85% discounts. AWS provides a two-minute interruption notice through instance metadata and CloudWatch Events. You can request spot capacity through EC2 Spot Fleet, Auto Scaling Groups, or directly through the EC2 console. Spot pricing varies by instance type, availability zone, and current demand.

AWS Spot Instance Advisor shows historical interruption frequencies by instance type and region. In the Mumbai region, general-purpose instances (m5, m6i) have relatively low interruption rates (5-10%), while GPU instances can face higher interruption rates during peak AI training periods.

Azure Spot VMs and GCP Preemptible VMs

Azure Spot VMs work similarly, with discounts of 60-80% in the Central India and South India regions. Azure provides a 30-second eviction notice. GCP offers Spot VMs (formerly Preemptible VMs) with 60-91% discounts in the Mumbai region. GCP Spot VMs can run indefinitely but may be interrupted at any time.

Each provider's spot pricing and interruption patterns differ. For Indian workloads, compare spot availability and pricing across providers for your specific instance types. Off-peak hours in India (11 PM to 8 AM IST) often offer better spot availability because domestic demand drops.

[IMAGE: Pricing comparison chart showing on-demand vs spot instance costs for common instance types in AWS Mumbai region - spot instance pricing India]

What Workloads Are Best Suited for Spot Instances?

Spot instances work for workloads that can tolerate interruptions. According to Gartner, 45% of enterprises use spot instances for batch processing and data analytics. The key characteristic is fault tolerance: if an instance disappears, the workload can resume or retry without data loss or service disruption.

Ideal Spot Workloads for Indian Enterprises

Batch processing jobs like report generation, data ETL pipelines, and bulk email delivery are natural fits. CI/CD pipelines that build and test code can restart from the last checkpoint if interrupted. Data analytics and machine learning training jobs that save progress periodically handle interruptions gracefully. Testing and QA environments don't need guaranteed uptime.

Indian IT services companies running test environments for client projects achieve massive savings with spot instances. A test environment running 10 m5.xlarge instances at on-demand pricing costs approximately INR 4.5 lakh per month. The same environment on spot instances costs INR 90,000 to INR 1.5 lakh, an immediate 65-80% reduction.

[ORIGINAL DATA]

Workloads to Avoid on Spot

Never run production databases, customer-facing web servers, or stateful applications on spot instances. These workloads can't tolerate two-minute interruption notices. API gateways, payment processing, and real-time streaming services need guaranteed capacity from reserved or on-demand instances.

A practical hybrid approach uses reserved instances for production baseline, on-demand for production bursts, and spot for everything else. This strategy captures spot savings where appropriate while protecting critical workloads from interruption.

[INTERNAL-LINK: ROI optimization → maximising ROI with cloud cost optimization]

How Do You Manage Spot Instance Interruptions?

Handling interruptions gracefully is the key to successful spot usage. According to AWS Compute Blog, teams that implement proper interruption handling achieve 95% cost savings compared to on-demand while maintaining workload reliability. The right architecture makes interruptions a non-event rather than a disruption.

Interruption Handling Strategies

Checkpointing saves work progress at regular intervals. If a spot instance is interrupted, the job resumes from the last checkpoint on a new instance. For Indian data processing pipelines, checkpoint every 5-10 minutes to minimise rework.

Diversification spreads workloads across multiple instance types and availability zones. AWS Spot Fleet and Auto Scaling Groups can request capacity from 10-15 different instance types simultaneously. If one type is interrupted, others continue running. This approach reduces the impact of any single interruption event.

Instance Diversification for Indian Regions

The Mumbai region offers a broad selection of instance types. Diversify across at least 5-6 instance families (m5, m6i, c5, c6i, r5, r6i) and all available availability zones. We've found that Indian enterprises using a single instance type experience 3-5x more interruptions than those diversifying across six or more types. Diversification is the simplest and most effective interruption mitigation strategy.

[CHART: Bar chart - Spot instance interruption rates by instance family in AWS Mumbai region - AWS Spot Instance Advisor 2025]

Frequently Asked Questions

Can I use spot instances for Kubernetes workloads?

Yes. EKS, AKS, and GKE all support spot or preemptible node pools. Run stateless workloads and batch jobs on spot node pools. Use on-demand node pools for stateful workloads and system components. Kubernetes' built-in rescheduling handles pod eviction when a spot node is interrupted, automatically moving pods to available nodes.

How stable is spot pricing in Indian regions?

Spot pricing in AWS Mumbai is relatively stable for common instance types (m5, c5), typically fluctuating within a 5-10% range. Less common instance types and GPU instances see more price volatility. Use Spot Instance pricing history in the AWS console to check 90-day pricing trends for your target instance types before committing to a spot-based architecture.

Are spot instances available in the Hyderabad region?

Yes. AWS launched the Hyderabad (ap-south-2) region with full spot instance support. Spot availability and pricing in Hyderabad are often slightly better than Mumbai because the region has lower overall demand. Consider spreading workloads across both Indian regions for better spot capacity access.

[INTERNAL-LINK: cost visibility → cloud cost visibility dashboard guide]