3 ms·
This Is What Twitter Does Not Want To Happen, Part Five
- ChuckMcM 15y agotl;dr version, twitter is returning an error for links that should exist, general whine about how unreliable it is. The meta comment here is that keeping a site running 24/7 of the scale of a Twitter or a Facebook or any number of high interaction rate businesses is not a 'simple' exercise. There isn't a lot of off the shelf software that can do that and while you can depend on AWS or another third party web 'stack' it comes with its own problems. I once pointed out to a principal in a cloud infrastructure company that while I understood that the concept of 'cloud' was that you didn't care what went on under the covers, if you didn't understand what went on under the covers you could make some really stupid decisions above the covers. I likened it to know what the compiler is doing when you build say a recursive tree walk. If you don't know how compilers work, and you decide to pass your tree node structures on the stack, you suddenly 'blow up' the stack when your tree gets above a certain (unpredicted!) size. Just like your cloud app can "blow up" when the kinds of things it is doing start exceeding the underlying platform infrastructure's ability to deliver. (network congestion or latency is usually in that mix).