Modern cloud computing didn’t emerge from nothing. The core concepts — virtualized resources, distributed scheduling, multi-tenancy, and on-demand provisioning — were developed and refined in academic grid computing research throughout the 1990s and 2000s.
What Was Grid Computing?
Grid computing connected heterogeneous computing resources — university clusters, supercomputers, research instruments — across wide-area networks to create a shared computational fabric. Resources were contributed by multiple institutions and shared through standardized middleware.
Key projects:
- Globus Toolkit — authentication, data transfer, and job submission across grid sites
- EGEE / EGI — European grid for particle physics and life sciences
- TeraGrid / XSEDE — US National Science Foundation high-performance computing grid
- GridWay — metascheduler developed at UCM that abstracted job submission across multiple grid sites
The Research-to-Cloud Pipeline
The Distributed Systems Architecture group at Universidad Complutense de Madrid worked at the intersection of grid computing and virtualization. Their work on scheduling virtual machines across distributed infrastructure led directly to OpenNebula, first released in 2008 — the same year AWS EC2 left beta.
Key concepts that transferred from grid to cloud:
| Grid concept | Cloud equivalent |
|---|---|
| Virtual Organization | Cloud tenant / account |
| Grid job scheduler | Cloud autoscaler / orchestrator |
| Data Grid | Object storage (S3, Blob) |
| Pilot jobs | Spot / preemptible instances |
| VO membership | IAM roles and policies |
| Site federation | Multi-region / multi-cloud |
What Changed
Grid computing required trust relationships between institutions and manual brokering of resources. Cloud computing solved this with:
- Self-service provisioning — no approval process; resources available in minutes
- Pay-as-you-go — commercial pricing replaced research allocation models
- Standardized APIs — consistent interfaces regardless of underlying hardware
- Virtualization at scale — hypervisors made multi-tenancy practical
The Legacy
Grid computing produced the scheduling algorithms, federation protocols, and resource management concepts that cloud providers implement at commercial scale. Researchers who built grid middleware in the 2000s — including the DSA group at UCM — became the engineers and architects who designed early cloud platforms.