Finally we have all the basics that are needed for building an Autocomplete service and now comes the juicy part. From now on we are looking at how to make it fast and robust. In the current approach we open a new Data base connection for every HTTP request. This needs quite some time to lock the data base (at least when using neo4j in the embedded mode) and then also to run the query without having any opportunities to use the caching strategy of the data base.
In this tutorial I will introduce you to the concept of a ContextListener. This is roughly spoken a way of storing objects in the Java Servlet global memory using key value pairs. Once we understand this the roadmap is very clear. We can store objects like data base connections or search indices in the memory of our web server. As from what I currently understand this could also be used to implement some server side caching. I did not do any benchmarking yet testing how fast retrieving objects from context works in tomcat. Also this method of caching does not scale horizontally well as using memcached.
Anyway have fun learning about the context listener.

If you have any suggestions, comments or thoughts or even know of some solid benchmarks about caching using the ServletContext (I did a quick web search for a view minutes and didn’t find any) feel free to contact me and discuss this!