Amazon, eBay, Google—the holy trinity of very large server rooms. Each wrestles with staggering transaction volumes, multiple data centres, and a never-ending stream of new services. In the quest for the perfect never-fail architecture, these companies are constantly pushing the boundaries of network science. Lately, they’ve started to share some secrets, at least of the non-technical, high-level variety. Amazon’s latest article in the Symposium on Operating Systems Principles is especially enlightening if you’re thinking about whipping together a massively decentralized, globe-spanning collection of server farms…
“customers should be able to view and add items to their shopping cart even if disks are failing, network routes are flapping, or data centers are being destroyed by tornados.”
