At about 2pm Bangladesh time on Monday, a large part of the consumer internet stopped working, and the reason was a single sentence on Amazon’s status page: increased error rates for the DynamoDB endpoint in the US-East-1 region. DynamoDB is a database service; US-East-1 is Amazon Web Services’ oldest and largest data-centre region, in northern Virginia. Because an enormous number of companies keep their most important data there, an error in that one place rippled outward until Fortnite, Snapchat, Roblox and Clash of Clans stopped loading, Perplexity’s AI search went dark, Coinbase and Robinhood froze trading, Venmo and Chime could not move money, Lyft could not dispatch cars, Signal’s messages stalled, and in Britain Lloyds Bank, Bank of Scotland, Vodafone and BT reported problems. Amazon’s own shop, Prime Video and Alexa were affected too.
What happened
AWS said engineers were working on the fault within minutes and that customers might be unable even to open support cases during the incident. Perplexity chief executive Aravind Srinivas posted that his service was “down” and the root cause was AWS. Signal president Meredith Whittaker confirmed her platform was hit. Elon Musk noted that X was still up. Recovery began around 6pm Bangladesh time, according to the Telegraph; Canva, among others, warned that downloads might remain patchy for a while. Downdetector logged it as the largest internet disruption since the CrowdStrike software failure of July 2024, which grounded flights and shut hospital systems worldwide.
Why one region matters so much
US-East-1 is where AWS launched in 2006 and where many of its global control services still live. Companies choose it because it is cheapest, has the most services and is the default in every tutorial. The result is concentration: an outage there affects customers who have never heard of Virginia. Cloud providers offer multi-region designs that survive exactly this failure, but they cost more and are harder to build, and most companies decide the risk is acceptable until the afternoon it is not.
Why Bangladesh barely noticed
Prothom Alo reported no disruption to apps or websites used from Bangladesh, and that is instructive. Local services — bKash, Nagad, the mobile operators, most banks, the government portals — run on domestic data centres or on cloud regions in Singapore and Mumbai, not in Virginia, and the American consumer apps that failed are lightly used here. That is protection by accident rather than design. The startups now building on AWS from Dhaka mostly default to the same US-East-1 as everyone else, and the lesson of Monday is the one every cloud architect already knows and most budgets ignore: a region is a single point of failure, and the second region is the insurance premium.
Source: The Guardian, via Prothom Alo.
