3 ms·
FWIW, a few years back, I was averaging <5ms for a benchmark of live traffic (many thousands of queries per second) on Pinterest queries against their full data
by fizx 9y ago
FWIW, a few years back, I was averaging <5ms for a benchmark of live traffic (many thousands of queries per second) on Pinterest queries against their full dataset on a single EC2 box with a weeks worth of customization on top of Lucene.
The existing stuff isn't slow.
- markpapadakis 9y agoTrinity - depending on the execution mode - is over 100% faster for certain queries compared to Lucene, and Lucene is already very fast. It all comes down to the postings ling codecs, and the iterators design/impl anyway.