We use solr running on tomcat6 as the search server, resque to queue them up, and standard http requests to query the docs. We can index about 15 documents per second.