>>643016i work with systems like this. you don't, and i can tell because you invoked "computer science" rather than the relevant field of "distributed systems architecture". so, my little retarded monkey, allow me to spoonfeed you a little.
youtube is experiencing an extended partial outage of its distributed message queues. it explains literally everything we see: the defective view counter, the dropped chat messages (and 24-hour outage of member stream chats!), the flapping subscriber counts, and the amazing disappearing waiting room. all of these things need to be transported around the world quickly, and the one kind of technology that can do that is a distributed message queue.
have you worked with a distributed message queue? have you worked with a single-node message queue? hell, can you even pronounce "queue"?
now, your homework is to tell me how you would diagnose this given the relevant administrative access to google's systems. tell me what you'd look at and how you'd prove or disprove my theory. go on.
