FeatureCloud

How apps run across sites and how a run is coordinated.

Status of this page

To be filled.

This page will describe the platform CVDLink uses to execute analyses across participating sites without moving patient data between them.

Intended contents

What FeatureCloud provides. The problem it solves, and how it relates to the rest of CVDLink.

Architecture. The controller and the participating clients, what runs at each site, and which parts are shared.

How a run is coordinated. How sites are invited and join, how the controller tracks progress, and what passes between sites during a run. Specifically, what is exchanged and what never leaves a site.

Installation. What each site installs, the ports involved, and how a site is registered.

Starting a run. Who initiates it, what the other sites see, and the point at which a run becomes active.

Monitoring. Where the state of a run is visible, and what the states mean.

Failure handling. What happens when a site drops out mid-run, whether a run can resume, and how to recover.

Relationship to the P2P component. CVDLink has its own federation service for metadata exchange. This page should make clear which responsibilities belong to which, so that readers do not confuse the two.