> For starters, we built the entire Bing stack from scratch. No OSS. No common platforms like the JVM. Nothing like that. We started from nothing, and invented the server infrastructure, the data pipeline, the runtime that would support the site, the ML tools, everything. The fact that the site runs at all is a small miracle, but the site does not "just" run: the most remarkable thing by far is that the quality of our tooling is quite incredible, generally an order of magnitude better than the OSS equivalents. For example, the largest deployment of an OSS NoSQL datastore seems to be a few thousand nodes. The small NoSQL cluster backing our MapReduce implementation is stably deployed on a cluster an order of magnitude larger than this. This is something you only really see at companies like Amazon, Google, or MS.
You lost me here. Not building on OSS seems like setting yourself up or failure from the start, particularly when you are fighting a manpower war, which is where OSS is beating every proprietary entity. OSS already powers Google and Amazon and OSS db's will scale to billions of nodes, not a few thousand.
You lost me here. Not building on OSS seems like setting yourself up or failure from the start, particularly when you are fighting a manpower war, which is where OSS is beating every proprietary entity. OSS already powers Google and Amazon and OSS db's will scale to billions of nodes, not a few thousand.