Just last month I wrote Single Points, about the Fastly CDN outage. This morning many, many sites were down or inaccessible because of an outage at Akamai. A content delivery network again, though they’re saying the outage is caused by ‘edge DNS.’ I’m familiar with DNS, but not the ‘edge’ variant. In fact, I realise it’s capitalised and is the name of an Akamai product or service.
More evidence that the increasing centralisation of internet services is a problem. On the plus side, it was resolved quickly. When a service provider has the kind of major clients we’re talking about here, then that company is going to have to be able to respond quickly and get things back up. If a random small or midlevel company ran all its own server hardware and software, an outage would only inconvenience that company’s customers. But the company would need to have the staff available to sort the problems out. That would be a large and arguably unnecessary overhead.
So I understand the desire to offload responsibilities to a service provider, and the economies of scale that a company specialising in running network services can bring. But I fear it’s only a matter of time before one of these events results in serious damage or even loss of life.
Not that I’m claiming to know what the answer is.
I noticed that GitHub was down this morning – or not down, exactly, but its web pages were profoundly broken. I tried different browsers, then jumped on Twitter to see if it was widely reported.
It was. People were saying the problem was Fastly, a content delivery network (CDN). Also that it was affecting other sites. I don’t know when CDNs started being a thing. I think they might have been recommended by some when I was still using WordPress. The idea being that a CDN can host your site’s static assets – images, mainly – while WordPress carries on with the dynamic bits, generating HTML pages on the fly, as it does. The CDN’s scale will mean it can serve those files faster than your little server.
I didn’t bother with them, not having that much traffic. But in the back of my mind there was always the thought, ‘What if the CDN goes down?’ The idea, of course, was that the CDN would be big, multiply-redundant, reliable: it’s not going to go down!
Here’s a CNN report about the outage. It affected a lot more than GitHub, it seems.
So, are CDNs single points of failure? Obviously there’s more than one CDN, but if the failure of any one can disable large chunks of the web, do they put us in a better position?
"The whole passport design is totally brain damaged," Mr Grunwald told Wired.com. "From my point of view all of these [biometric] passports are a huge waste of money - they're not increasing security at all."
No surprises there, then. Except maybe how quickly it’s happened. Single point of failure, anyone?