Low-latency information retrieval for heavy-read/low-write content sets. An example that might solidify the idea: pack a Wikipedia snapshot into it for search,

Low-latency information retrieval for heavy-read/low-write content sets. An example that might solidify the idea: pack a Wikipedia snapshot into it for search, and serve ~1m queries per second on it (12k qps per core).