About¶
Vishwanath Grid is a research-computing cluster that several Indian universities run together. A "cluster" in this context just means a group of computers, sitting in a server room, that researchers use to run scientific work — simulations, data analysis, machine-learning training, anything that takes longer than a laptop can handle.
What makes this one different: instead of every university buying its own cluster and keeping it locked to its own people, the partner sites share. A researcher signs in once, submits a job, and the work goes to whichever site has free capacity that day.
Everything that runs the cluster is open-source software. There are no paid subscriptions in the stack that a vendor could turn off later.
Mission¶
To put production-grade research computing within reach of every serious Indian university — without making any of them dependent on a vendor, a cloud provider, or a single funding stream.
Vision¶
A federation of Indian universities running shared, open-source research infrastructure on hardware they own and operate — sized modestly today, growing as more partners join, and durable enough to outlive any single grant or political cycle.
Operating principles¶
- Open by default. Every piece of software in the stack ships under a permissive open-source licence. The configuration that runs the cluster is also open and version-controlled in the public Forgejo repository.
- Partner-owned hardware. Each partner site keeps physical and administrative control of its own machines. Nothing routes through a central cloud provider; no partner is locked in.
- No commercial workloads. The cluster is for academic research, reproducibility, and teaching. Cryptocurrency mining, commercial ML inference, and similar workloads are explicitly out of scope — see Governance.
- Honest about the stage. The project is early. Phase 0 is one pilot site. We say so on every page and we mean it.
- Boring infrastructure choices. Where two technical options are roughly equivalent, we pick the one that has been in production at academic facilities for ten years. HTCondor (CERN, NASA HEC), Debian, Keycloak, Ansible — not the latest framework of the month.
Inspiration¶
The project is named for and inspired by Dr. Vishwanath D. Karad, founder of MIT World Peace University, Pune and the World Peace Dome — a lifelong proponent of education as a public good and of "union of science and spirituality" as a frame for what universities are for. The federation model also draws on three established academic templates: the Worldwide LHC Computing Grid at CERN, the Open Science Grid in the United States, and the EuroHPC Joint Undertaking — all of which are federations of independent partner sites that look, from the outside, like one shared facility.
Why we are building this¶
Indian research labs hit a compute wall every year. Three things happen at once:
- Commercial clouds get expensive fast for the kinds of long-running scientific workloads researchers actually have.
- Buying a private cluster is feasible, but keeping it running needs specialist staff most departments cannot hire.
- The national HPC facilities have long queues.
The result is that a lot of useful work either waits, or quietly moves to overseas cloud providers. We can do better than that with the hardware Indian universities already have on campus — if it is pooled.
What "federated" means here¶
A federated cluster means the computers are physically spread across multiple sites, but a researcher sees one cluster. The software that schedules work — what we call the scheduler — knows where every node is and routes each job to a free machine.
Each partner site keeps full control of its own hardware. Nothing is moved off-campus. Partners can audit every job that runs on their machines, and they can leave the federation at any time without losing their hardware or their data.
Where we are today¶
| Phase | What it is | Status |
|---|---|---|
| 0 | Pilot at MIT-WPU Pune, scheduler running, researchers submitting jobs | Live |
| 1 | Two more Indian universities joining as partner sites | In progress |
| 2 | The rest of the everyday-research toolkit, behind one login | Planned |
For the detailed version, see the roadmap.
Who is building it¶
The current team is the research-computing group at MIT World Peace University, Pune. The intent is that, over time, every partner university contributes operators of its own.
A note on tone¶
This site is the public-facing snapshot of an early project. We are matter-of-fact about what is live (the Phase 0 pilot) and what is not (everything past Phase 1). If something here is unclear, ask us.