Where studies will run

No GPU to own and nothing to install. Connecting the machines you do have, and choosing for each study where it runs, is planned.

MDXplora is being built and opens to early-access groups in stages. This site describes it as planned.

Why run studies on MDXplora

Nothing to install

A browser will be enough: no environment to build and no drivers to match to a CUDA version.

The same study anywhere

A study is one Config. It will run with the same settings on our GPUs and on yours, and it can be downloaded whole, with its data and its record, whenever you want it.

Chosen per study

Our GPUs when you have none, and your own workstation or cluster when you do, once the connector is ready.

Connecting your own machines

The connector is not available yet. It is designed so that your machine asks MDXplora for work, the way a mail app checks for new mail. Nothing reaches into your network, so there are no ports to open. As designed:

  1. 1

    Install the connector, once

    One command on your workstation or on your cluster’s login node, then sign in with your MDXplora account.

  2. 2

    Your machine appears in MDXplora

    It shows up under “Run on”, beside MDXplora’s own GPUs, with what it has: its GPUs and, on a cluster, its scheduler.

  3. 3

    Choose where each study runs

    Build the study in the browser, pick the machine and press Run. Each choice shows the estimated time before you commit.

  4. 4

    The connector does the work

    It picks the job up, runs the study there or submits it to SLURM, and sends progress back to your browser.

Your data can stay where it is

Trajectories will be able to remain on your own storage, with only the study records and reports shown in MDXplora.

It only runs MD studies

The connector is designed to run validated Configs and nothing else, never code sent to it, so an IT department can see exactly what it does before approving it.

Ask for early access

Early access is opening to a small number of groups first. Tell us about your systems and your compute, and we will be in touch.

Join early access