REACH

High-performance computing for Chicago researchers who don’t have any.

REACH runs a shared cluster built from donated enterprise servers, so researchers at smaller universities can run genomics, machine learning, and simulation work without a cloud bill or a data center of their own.

Donate hardware Run your research here

Chicago, Illinois. A nonprofit in formation.

Illustration of a server rackA rack of five mismatched servers and network gear with an empty section at the bottom where more donated hardware will go.
Illustration of REACH’s first rack: mismatched donated servers, one scheduler, and room left for more.

Big universities run their own clusters. Smaller schools mostly can’t.

Faculty and students at smaller institutions end up running large jobs on laptops, waiting on borrowed accounts, or spending grant money on cloud computing. Meanwhile, companies retire capable servers from their data centers every year.

REACH connects the two. We take in donated hardware, build it into a shared cluster, and give researchers access at low or no cost. The full name is Research Equity and Access Computing Hardware.

How it works

  1. Hardware is donated

    Companies and institutions send us servers, storage, and networking gear that they are retiring.

  2. We sort it

    Machines that can do useful science join the cluster. The rest are refurbished and resold to fund operations, or recycled through certified partners.

  3. Researchers compute

    Approved researchers log in through a web portal, submit jobs, and collect their results. No servers to manage, and no need to be a systems expert.

Where the build stands

Servers, racked and networked
Four servers and a managed switch, with remote power and console control on every machine.
Built
Large-memory node
One server with 1.2 TB of RAM, for genome assembly and other datasets that will not fit on a laptop or a standard cloud instance.
Built
Job scheduler
Slurm, the scheduler used by most research clusters.
Underway
Web portal and sign-in
Browser access through Open OnDemand, with multi-factor authentication.
Planned
High-speed interconnect and shared storage
InfiniBand between nodes and a shared file system, so jobs can span machines.
Planned
GPU nodes
Graphics cards with large memory for machine learning and genomics workloads.
Planned

Looking for our first labs.

We are opening to a small first group of researchers. If you work at a Chicagoland college or university and your work has outgrown your machines, tell us what you run.

Useful details: the software you use, how large your data is, and how long a typical job takes.

Email us about your work