14 DAY TRIAL // CloudFlare, despite being only a few years old, has become quite popular. So popular in fact that, when it goes down, it takes hundreds of thousands of sites with it, something not even Amazon can boast about.
Granted, a CloudFlare outage is not the same as an Amazon Web Services outage, but it's a problem for a lot more websites as well.
CloudFlare acts as a CDN but also provides some security features, such as protection from DDoS attacks, thanks to the distributed nature of the service.
In fact, the outage was the result of the company fighting back against a DDoS attack on one of its clients.
Its experts created a new routing rule to counter the attack and then pushed it to CloudFlare's edge routers, the ones facing the internet.
However, instead of the routers accepting the rule and continuing to operate, they started using up all their RAM until they crashed, so hard that they had to be manually rebooted and have the rule reverted at the same time.
CloudFlare explained that it took no more than 30 minutes for the first routers to come back online, but then all of the traffic got directed to the few working routers causing more problems. As more routers came back online, things got back to normal.
CloudFlare is now offering a refund to its customers and has contacted the router manufacturer, Juniper, to determine the cause of the problem.
One of the reasons why the web has succeeded and why it's so reliable is that it's a distributed system, there are trillions of pages hosted by hundreds of millions of servers. If one fails, even if 1,000 fail, there's no problem.
While the web and the internet continue to grow, there is a huge trend towards centralization, fewer big sites and fewer platforms or tools hosting them, like Amazon Web Services and its competitors, CDNs or CloudFlare.