Improve performance of Ajax User Picker

http://confluence.atlassian.com/display/JIRA/JIRA+3.10+Release+Notes#JIRA3.10ReleaseNotes-ajax

We are really excited by Ajax User Picker and want to use it in our environment.
But the problem is that we have more than 5K of users ( infact more than 50K ).

Current implementation of Ajax "User Picker" is very CPU greedy, because it does simple sequential scan ( without any index access ) of all jira users.

Here's the code:
com.atlassian.jira.bc.user.search.DefaultUserPickerSearchService

final Collection users = getAllUsers();
for (Iterator iterator = users.iterator(); iterator.hasNext();) {
    final User user = (User) iterator.next();
    if (userMatches(user, query, canShowEmailAddresses)) {
        returnUsers.add(user);
    }
}

It makes sense to improve performance of such operation by utilizing either

• Lucene index + Prefix query
• Database index on id / firstname / lastname