With Siri getting asked hundreds of thousands of questions everyday, it must take a lot of computer processing, considering it has to make out the words from different accents across the globe, analyze it, process it, and then return a witty and appropriate response, all in ~2 seconds.
But how does Apple manage all this? The answer comes to us in the form of a Reddit post, by the ex-colleague of Apple’s Lead Cloud Architect. According to him, there are 3 Siri ‘server farms’ in the United States, with at least one farm for every other country out there.
He explains that each server farm has 4 HP c7k enclosures, with 8 HP server blades each, with up to 1TB of RAM each. In total, that gives each server farm or ‘instance’ 32 HP servers with a whopping 1024 cores, and 32TB of RAM.
Adding up the costs of the enclosures and the servers, we get $1,019,008 per server farm, excluding any other expenses.
Check out the original post over at Reddit here.