Ask HN: How often is hardware at the big tech companies refreshed?
Does anyone have data that would suggest how long it takes an org at the scale of an Amazon, Google, or Facebook to entirely replace their HW?
I know that current practice at that scale is to not replace an individual chassis, only whole racks when a certain % of servers have failed, but that can't be the whole story, since presumably servers/racks are replaced if they are old enough even if they happen to still be running OK.
Google for example almost certainly has no servers (or switches, etc.) running that date back to Y2k even though statistically, some 16-year-old HW would still be working if it hadn't been yanked.
So, assuming the HW hasn't actually failed, how old is too old for a "web-scale" company to keep around? And is the answer different for these companies' own services vs. externally facing IaaS/PaaS?
I know that current practice at that scale is to not replace an individual chassis, only whole racks when a certain % of servers have failed, but that can't be the whole story, since presumably servers/racks are replaced if they are old enough even if they happen to still be running OK.
Google for example almost certainly has no servers (or switches, etc.) running that date back to Y2k even though statistically, some 16-year-old HW would still be working if it hadn't been yanked.
So, assuming the HW hasn't actually failed, how old is too old for a "web-scale" company to keep around? And is the answer different for these companies' own services vs. externally facing IaaS/PaaS?