I am planning to search large(ish) data sets and need to get results back quickly (a few ms). I have an application I am working on which uses a relational database, mainly for integrity of the information, so I am not planning to change this engine.

However I will need to factor in stuff like the popularity of the result and relevance. I would rather not lather my queries results with code and would much rather pick a good tool to do this searching. e.g. Im talking about typing a single character like 's' for example, I would want to get an autocomplete for the most popular items that start with 's' and so on. I see NoSQL can provide this information really really quickly. I don't mind keeping a score of the most popular results with my own code if that information is needed, but if there are engines that do this already then their algorithms will probably be better refined.

In the past I have been able to use google I am looking at elastic search and other no-sql products. Has anyone ever done something like this or has used a product that has done it quite well? I am not gonna jump in and use it off the bat, but I might set up a copy or a trial and see if its useful or not.