"No realistic alternative": Signal boss defends use of AWS

The president of the foundation responsible for Signal finds it worrying that so many people were surprised that the popular messenger was affected by the major AWS outage. Meredith Whittaker explained this in several posts on Mastodon and Bluesky, in which she explains why the software would not work without Amazon's infrastructure. The immense concentration of power in the hands of a few so-called hyperscalers is much less understood than she had assumed. Nevertheless, her explanation for the great dependence on Amazon, Google, Microsoft, and Google has at least some gaps.

When the disruptions at AWS from Amazon began a week ago, Signal was one of the first services that struggled with disruptions. For more than an hour, sending and receiving messages was prevented. This apparently caused a lot of surprise and criticism; Whittaker now indicates that many users were concerned about their chat content, which is routed through Amazon's servers. That's why she once again begins her explanations with the assurance that neither the US company nor Signal itself has access to this content. Thanks to the encryption used, no one except the participants in a chat can view it.

Infrastructure distributed across the globe necessary

The fact that Signal is in principle dependent on the hyperscaler AWS's infrastructure is due to the requirements of a “global platform for mass real-time communication,” she explains. To be able to transmit audio and video calls with low latency, it requires “a pre-built, planet-wide network of computing, storage, and edge presence.” Such infrastructure requires constant maintenance, consumes considerable electricity, and demands continuous attention and monitoring. For the requirements of instant messaging, there is no alternative to AWS & Co. She also clarifies that this is about more than just “renting a server.”

Such infrastructure costs “billions upon billions of US dollars,” writes Whittaker, which is why neither Signal nor others can build something like this. However, she also claims that even the hyperscalers themselves have to cross-subsidize this business. The cloud divisions of Amazon, Microsoft, and the like are among their most profitable business areas. Whittaker further writes that for the reasons mentioned, almost all services that offer real-time services are at least partially dependent on hyperscalers. She explicitly mentions Mastodon, but the microblogging service is decentralized and if at all dependent on other, smaller providers. The significantly smaller Signal competitor Threema, according to its statement, requires even only two servers in Switzerland.

AWS outage with unexpected consequences

Amazon itself explained the outage at the end of last week and attributed it to a single point of failure. As a result, numerous internet services could not be used or only with restrictions; for example, Amazon's video services also had issues. There were also problems with the Epic Games Store and Fortnite, Duolingo, some Apple services, Atlassian, and Docker, as well as Perplexity's AI technology. Particularly sensational were the disruptions with networked mattresses, which caused the control of the devices to fail. For Signal, the disruptions were resolved after an hour and a half.

(mho)