"We have been adding new servers and extra capacity at a record pace every single day this week nonstop, but today exceeded even our most optimistic projections. Millions upon millions of new users are sending a message that privacy matters. We appreciate your patience."
can't wait for the post mortem : o really wondering why there's an outage. I assume the SGX stuff quickly became a bottleneck, or the message forwarding server wasn't made to scale out very efficiently. In any case, they should have been able to have a very large queue in front of these components and just have a very slow system, probably something blew up due to the sudden peak of users.
Does anyone outside Signal Foundation know how's their architecture? There are a lot of references to AWS, GCP, and Azure in the source code hosted in GitHub so they probably use them all in one way or another. It would be super interesting know more details about the infrastructure.
interesting but doesn't really explain much. Is it really a single Signal server or can you spin many? What is stored on redis and postgresql? Maybe redis is used as cache, or as a queue, or as a store for sessions, and postgresql is used for accounts. Are these components distributed as well?
Despite recent events, I don’t see this happening. Signal has a strong reputation as a privacy-supporting app without any weird fringe Internet subculture attached to it. And I’m unaware of any public groups like you find on Facebook or WhatsApp.
De-platforming a private messenger like Signal would be a strong signal against an open, unmonitored Internet and raise too many eyebrows. Amazon would be seen as complicit with government for not allowing citizens to communicate freely and privately with each-other.
If Signal adds features to let users post extremist views publicly, that’s a whole different ballgame and the cards are off the table.
https://twitter.com/signalapp/status/1350165610936766464