Uber
1 article in the last 30 days.
-
Uber Cuts Translation Cache Memory by up to 90% While Keeping P99 Latency Below 1 MS
Uber's hybrid Go translation cache keeps keys and hot values in memory and the rest on disk. It cuts initial memory use by over 90%, keeps hit rates above 99.9% and holds p99 disk-miss latency under 100 µs.