This repo contains only always free resources from various public clouds. Contributions welcome!
Inspired by cloudcommunity/Cloud-Free-Tier-Comparison
WARNING I do not guarantee for any correctness of this list. It might be outdated (though I will try my best to keep it updated). It might also happen, that you charged for additional cost like traffic, etc.
All constraints (e.g. traffic) are tied to a period of one month.
Resource | Cloud | Product Name | Info |
---|---|---|---|
Database | Alibaba | PolarDB | 50GB SQL |
AWS | DynamoDB | 25GB NoSQL | |
Azure | Cosmos DB | 25GB NoSQL | |
IBM | Cloudant | 1GB NoSQL | |
IBM | Db2 | 0,2GB SQL | |
Oracle | Autonomous JSON Database | 20GB SQL | |
Oracle | Autonomous Transaction Processing | 20GB SQL | |
Oracle | NoSQL Database | 25GB NoSQL | |
render | PostgreSQL | 1GB SQL | |
render | Redis | 1GB NoSQL | |
S3 | GCP | Cloud Storage | 5GB only USA, region: us-west-1 OR us-central1 OR us-east-1, 1GB external traffic to every region except China or Australia, incomming traffic free |
IBM | Object Storage | 25GB | |
Oracle | Archive Storage | 10GB | |
Oracle | Object Storage | 10GB | |
Oracle | Object Storage - Infrequet Access | 10GB | |
Serverless | Alibaba | Function Compute | 1 Million Requests |
AWS | Lambda | 1 Million Requests | |
Azure | Container Apps | 2 Million Requests | |
Azure | Functions | 1 Million Requests | |
GCP | Cloud Functions | 2 Million Requests, 5GB external traffic to every region except Australia, incomming traffic free | |
GCP | Cloud Run | 2 Million Requests, 1GB external traffic to every region, incomming traffic free | |
VM | GCP | Compute Engine | e2-micro, provisioning model: regular, boot disk type: Standard persistent disk, region: us-west-1 OR us-central1 OR us-east-1, boot disk size: 30GB, 1GB external traffic to every region except China or Australia, incomming traffic free |
Oracle | Compute | 2 VMs with 1/8 OCPU and 1GB of memory each - 1OCPU = 2vCPUs -> 2x (0,25vCPU, 1GB RAM), 200GB Storage total | |
Oracle | Compute (ARM) | ARM VM(s) with 4vCPUs, 24GB RAM, 200GB Storage total -> e.g. 4vCPUs, 24GB RAM, 200GB Storage OR 2x (2vCPUs, 12GB, 100GB Storage) |
Cloud | Resource | Product Name | Info |
---|---|---|---|
Alibaba | Database | PolarDB | 50GB SQL |
Serverless | Function Compute | 1 Million Requests | |
AWS | Database | DynamoDB | 25GB NoSQL |
Serverless | Lambda | 1 Million Requests | |
Azure | Database | Cosmos DB | 25GB NoSQL |
Serverless | Container Apps | 2 Million Requests | |
Serverless | Functions | 1 Million Requests | |
GCP | S3 | Cloud Storage | 5GB only USA, region: us-west-1 OR us-central1 OR us-east-1, 1GB external traffic to every region except China or Australia, incomming traffic free |
Serverless | Cloud Functions | 2 Million Requests, 5GB external traffic to every region except Australia, incomming traffic free | |
Serverless | Cloud Run | 2 Million Requests, 1GB external traffic to every region, incomming traffic free | |
VM | Compute Engine | e2-micro (provisioning model: regular, boot disk type: Standard persistent disk, region: us-west-1 OR us-central1 OR us-east-1, boot disk size: 30GB, 1GB external traffic to every region except China or Australia, incomming traffic free | |
IBM | Database | Cloudant | 1GB NoSQL |
Database | Db2 | 0,2GB SQL | |
S3 | Object Storage | 25GB | |
Oracle | VM | Compute | 2 VMs with 1/8 OCPU and 1GB of memory each - 1OCPU = 2vCPUs -> 2x (0,25vCPU, 1GB RAM), 200GB Storage total |
VM | Compute (ARM) | ARM VM(s) with 4vCPUs, 24GB RAM, 200GB Storage total -> e.g. 4vCPUs, 24GB RAM, 200GB Storage OR 2x (2vCPUs, 12GB, 100GB Storage) | |
S3 | Archive Storage | 10GB | |
S3 | Object Storage | 10GB | |
S3 | Object Storage - Infrequet Access | 10GB | |
Database | Autonomous JSON Database | 20GB SQL | |
Database | Autonomous Transaction Processing | 20GB SQL | |
Database | NoSQL Database | 25GB NoSQL | |
render | Database | PostgreSQL | 1GB SQL |
Database | Redis | 1GB NoSQL |
You'll also find some hints for gaining and keeping the access to those resources (e.g. by migrating your "trial" account to a paid one).
And if that's not enough, you might also find one or another example on how to use those free resources efficiently and maybe tie them together.