by John M Costa, III

Kuberentes Hosting Services


When looking for a hosting platform for Kubernetes, I wanted to find a platform which was easy to use, had a good developer experience, and that was cost-effective. Easy to use is somewhat subjective and certainly depends on familiarity with the platform, domain knowledge, and other factors. Therefor, I’ll try to be as objective as possible when evaluating the platforms looking at Developer Experience and Cost Effectiveness.

For others, there could be other dimensions which are more important. For example, if you’re looking to meet certain compliance requirements, you might want to look at the security and compliance features of the platform and rate them accordingly.

For me and my project, these are not yet significant concerns.

Hosting Platform Options

An AI Assisted search via OpenAI’s ChatGPT1 for Kubernetes hosting platforms yields the following results:

Hosting ProviderCost EffectivenessDeveloper Experience
AWS- Components: EC2, S3, RDS, Lambda, etc.
- Pricing: Pay-as-you-go model, variable costs
- Productivity: High
- Impact: Broad range of services
- Satisfaction: Generally positive
Google Cloud- Components: Compute Engine, Cloud Storage, BigQuery, etc.
- Pricing: Sustained use discounts, per-minute billing
- Productivity: High
- Impact: Advanced AI and ML capabilities
- Satisfaction: Positive developer tools
DigitalOcean- Components: Droplets, Spaces, Databases, etc.
- Pricing: Simple and transparent pricing, fixed monthly costs
- Productivity: Moderate (simplified services)
- Impact: Suitable for smaller projects
- Satisfaction: Good user interface
Azure- Components: Virtual Machines, Blob Storage, Azure SQL Database, etc.
- Pricing: Flexible pricing options, Hybrid Benefit for Windows Server
- Productivity: High
- Impact: Integration with Microsoft products
- Satisfaction: Depends on familiarity with Microsoft ecosystem


create a markdown table which includes the following hosting providers:
Google Cloud

use the following columns so that each option could be evaluated:
- developer experience
- cost effectiveness

developer experience should include productivity, impact, satisfaction
cost effectiveness should include components and pricing for those components

Validating the Findings

Cost Effectiveness

The following are specifications for a development environment. The goal is to have a non-high availablilty Kubernetes cluster with 2 worker nodes intended for a development environment. The cluster should have a managed control plane and managed worker nodes, and should have object storage and load balancing. The cluster should also have a managed Kafka instance.

Pricing has been calculated generally using two worker nodes, and the cheapest option for the managed control plane.

Monthly Pricing (as of November 20232):

AspectAWS3Google Cloud4DigitalOcean5Azure6
Managed Control Plane73.00 USD73.00 USD00.00 USD73.00 USD
Managed Worker Nodes27.45 USD97.09 USD36.00 USD175.20 USD
Object Storage00.02 USD0.023 USD05.00 USD52.41 USD
Load Balancing31.03 USD18.27 USD12.00 USD23.25 USD
Managed Kafka86.58 USD31.13 USD15.00 USD10.95 USD
Managed Database69.15 USD25.55 USD15.00 USD24.82 USD
Total287.23 USD245.97 USD83.00 USD359.63 USD

Developer Experience

GitHub mentions Developer Experience (DevEx)7 as productivity, impact, and satisfaction. My thought is to document my experience so that other’s can evaluate the platforms for themselves.

Given the pricing schedule above, it’s not currently feasible for me to fully evaluate all the platforms at the same time. Instead, I’ll focus on the most cost-effective one, DigitalOcean. If given the opportunity and necessity, I’ll evaluate the other platforms in the future.

In a follow-up article, I’ll report my observations and experience. For now, I’ll leave this as a placeholder.

Thanks for reading!

  1. ↩︎

  2. For expediency, I’ve tried to choose similar services across the platforms. A better evaluation might detail the precise specifications of each service. For expediency, I’ve chosen to leave out some of these details and could backfill them if they became more relevant. ↩︎

  3. 2 t3a.small nodes as workers ↩︎

  4. 2 n1-standard-2 nodes as workers ↩︎

  5. 2 Standard Droplets as workers, 1GB of object storage ↩︎


  7. ↩︎

comments powered by Disqus