How I Used OpenStack APIs to Build a PaaS for ATM Mobilis

This article focuses strictly on the infrastructure and cloud orchestration layer of a Platform as a Service (PaaS) system we designed for ATM Mobilis using OpenStack APIs. It does not cover web development or application backend implementation. The focus is entirely on cloud infrastructure design, provisioning, and automation.

Context

The goal of the project was to design a PaaS layer that abstracts cloud infrastructure complexity and allows controlled provisioning of compute, networking, and storage resources on demand.

We built this system on top of Openstack leveraging its APIs as the core infrastructure control plane.

What we built on top of OpenStack

Instead of interacting manually with cloud resources, we built a control layer that orchestrates:

Compute provisioning
Network configuration
Storage allocation
Resource lifecycle management

All infrastructure operations were executed through OpenStack APIs.

OpenStack API layer usage

We relied on the core OpenStack services:

Nova for compute provisioning
Neutron for networking configuration
Cinder for block storage management
Keystone for authentication and service identity

This allowed us to treat infrastructure as programmable resources rather than manual configurations.

Infrastructure orchestration workflow

The platform follows a structured provisioning pipeline:

Request → Resource validation → OpenStack API calls → Provisioned infrastructure → Status tracking

Each request triggers a controlled sequence of OpenStack operations to ensure consistent infrastructure deployment.

Compute provisioning (Nova)

We used the compute API to dynamically provision virtual machines based on workload requirements.

Key operations included:

VM creation
Flavor selection
Image management
Instance lifecycle control

Nova API → VM instance → Assigned compute resources

Networking layer (Neutron)

Networking was fully abstracted using OpenStack networking APIs.

We handled:

Private network creation
Subnet allocation
Floating IP assignment
Security group configuration

This ensured isolated and controlled communication between deployed resources.

Storage management (Cinder)

Storage volumes were dynamically attached to compute instances.

We used:

Volume creation
Attachment to instances
Persistent storage lifecycle management

This allowed workloads to maintain state independently of compute instances.

Resource lifecycle management

A key part of the system was managing the full lifecycle of resources:

Provisioning
Monitoring state
Scaling operations
Decommissioning unused resources

This ensured infrastructure efficiency and prevented resource waste during operation.

Key engineering challenges

During implementation, the main challenges were not application-level, but infrastructure-level:

Ensuring consistent API orchestration across services
Managing dependency ordering between compute, network, and storage
Handling partial failures during provisioning
Maintaining consistent state across distributed resources

Conclusion

This project demonstrated how OpenStack APIs can be used as a foundation for building a PaaS layer that abstracts infrastructure complexity while maintaining fine-grained control over compute, networking, and storage resources.

The focus was entirely on cloud infrastructure orchestration and distributed resource management, not on application or backend design.

I worked on this project with my team as part of a real-world deployment scenario for ATM Mobilis, focusing on building reliable and scalable infrastructure automation on top of OpenStack.