A Compute Platform for Crisis Response

Cloudflare is extending its Project Galileo protections to cover the Workers compute platform for web projects related to the COVID-19 pandemic. Project Galileo was originally designed to shield at-risk projects from cyber attacks; now, the company is broadening that mandate to provide free access to Workers infrastructure for sites, applications, APIs, and tools responding to the public health emergency.

Workers offers a compelling technical proposition for fast-moving crisis projects: there is no underlying infrastructure to scale or maintain. A project can move from code to production in minutes, and traffic spikes are absorbed automatically by Cloudflare’s edge network. That fits the profile of several initiatives that have already sprung up during the crisis.

Waiving Workers Fees through Project Galileo

This move follows Cloudflare’s earlier decision to make Cloudflare for Teams free for small businesses through at least September 1, 2020, in response to the sudden demand for remote-access tooling. Developers working on COVID-19 related projects can submit a request through the Project Galileo application page.

Case Studies: Workers in the Field

Several developers have shared how they used Workers to stand up public-facing services during the pandemic. Their projects illustrate different approaches to leveraging edge compute for data aggregation, static hosting, and dynamic routing.

API-COVID19-In

Built by Amod Malviya, this API tracks COVID-19 cases in India using data from the Ministry of Health and Family Welfare alongside unofficial sources. Malviya designed it so teams across the country could build transparency tools, detect patterns in patient data, and plan hospital capacity off a common data layer.

"I created api-covid19-in to make it easier for people working all over India to contribute to fighting this situation — be it by creating mass transparency (the aggregate data API), or detecting patterns (the crowd sourced patient data API), or planning (hospital beds API)".

Malviya chose Workers because the code went from first write to live service in under an hour. That beat maintaining an origin server, which carried higher cost, and GitHub Pages, which couldn’t perform compute on each request. The lack of scaling or performance concerns was a decisive factor.

MakeFaceMasks

This project emerged from a Belgian grassroots makers’ movement and published a DIY mask-sewing manual approved by the Belgian Government. The website’s deployment flow relies on automated builds: GitHub Actions triggers Workers Sites deploys whenever content changes, which also simplified the process of generating translated versions of the site.

The load profile made Workers a natural fit. On launch day, the site absorbed 100,000 unique visitors without downtime, and the team did not have to tend to infrastructure during that traffic surge.

Mask a Hero NY

A volunteer-run site that connects medical professionals in need of personal protective equipment with donors. The project coordinated over 27,000 mask donations and was covered by NBC News, CBS, MSNBC, and other outlets after launching in roughly two days.

"When we built the site, we wanted the absolute easiest and most straightforward tech stack... the last thing I want to deal with on a static site is complex server configuration and uptime."

The site uses Jekyll as a static site generator and Workers Sites for hosting. A single wrangler publish command after a Jekyll build handles both deployment and cache-busting. Cloudflare’s built-in caching absorbed the traffic spikes that followed media features, freeing the volunteer team to focus on matching PPE with hospitals rather than watching server metrics.

CovidTracking API

The COVID Tracking Project publishes state-by-state testing data for the US, and its API is built on Workers as a serverless proxy. The data service initially synced with a Google Sheets spreadsheet, but the developer moved to Workers to normalize the source formatting and expose clean JSON.

An early version simply proxied the origin, creating an API request for every Worker request. To improve performance, the developer passed cf: { cacheEverything: true, cacheTtl: 120 } into the fetch() options so Cloudflare could cache origin results. However, some endpoints required parsing large XML payloads, which pushed CPU time past the 50ms limit and forced generous increases from Cloudflare’s side.

The solution was Cloudflare Key Value storage. Parsed and serialized results get saved to the KV store with a TTL of roughly an hour. On each incoming request, the previous cached result is returned immediately, then the Worker checks the item’s age. If it’s more than five minutes old, the Worker regenerates the payload and saves it for the next request. End users get fast, cached results, and the origin only gets hit when necessary.

"Google often requires an API key or has some strange formatting. I just wanted an array of values that reflected the sheet rows. No long complicated URL. I picked Cloudflare Workers because it works really well as a serverless proxy."

Resources for New Builders

Developers who want to contribute COVID-19 related tools but are new to Workers have several entry points:

  • Workers Sites: Deploy a static site directly to Cloudflare’s network with a few commands, using a tutorial or the accompanying video.
  • Localize a website: Translate your site to reach a wider audience during a global crisis.
  • Chat bot: Remote communication relies heavily on chat; building a bot can expose critical data in an accessible channel.
  • Template gallery: Boilerplate and code snippets, including an Apollo GraphQL server for API-style projects.
  • HTMLRewriter API: A streaming HTML parser available in the Workers runtime, useful for aggregating data from disperate sources; a working example is in the web scraper project.
  • The Built with Workers gallery can spark ideas for what type of application to launch.