I think you may need a much more advanced design - with change detection, parallel query execution, and index modification.
A lot of it depends on you semantics of a search - does it mean at the results are 'almost right' at a moment in time, or are pending index changes made first before any queries. How you define this will affect the architecture greatly. -----Original Message----- From: wenjie zheng [mailto:[EMAIL PROTECTED] Sent: Wednesday, April 12, 2006 8:45 PM To: java-dev@lucene.apache.org Subject: Question about RemoteSearchable, RMI and queries in parallel My question is a little bit long. Thanks for your patience. I am working on project which requires remote searching ability. So I use RMI and RemoteSearchable class. Here is the system structure: Server A has all the indices on it and the RemoteSearchable object running on it. Server B accepts queries and use ParallelMultiSearcher and RMI to search the indices on Server A. There are many indices on Server A, based on query I received, I choose different index to search from . For example, if I the query is from X, we will search indexX; if the query is from Y, we will search indexY. In order to do that, I changed the Searchable interface a little bit by adding function load(String folder) and implement it in all the subclasses. What load() does is to close the underlining reader of the first index and open the reader for the second index. So here is the problem I bumped into. When there are more than one queries coming at the same time and I wanted to process them in parallel, the load() function called by the second query may accidently close the the reader of the first index, while searching on the first index is not finished yet. So, do I have to queue the queries up and execute them one after the other? Or do I have other options? Thanks, Wenjie --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
