Hi, I came across this post, which contains interesting information about Lucene speed: http://www.5341.com/msg/2639.html
And yes, I noticed the date, it is indeed long time ago they posted the results. I have also built the 6M autocomplete/autosuggest index for the autocomple you saw last week and I can confirm it is very responsive, even when using word similarity for lookups - the worst response time was 1s, for a very long multi-token query. But usually, it was around 0.2-0.5s (index size was 2GB, read from disk, with limited RAM). I shall have numbers later, but the lucene autocomplete is *definitely* on the safe side concerning scalability and speed. Cheers, Roman
