Have you considered sharding the index using some logic? such as alphabetically or perhaps by document id etc. That way you have ||el indexing and searching.
On Wed, Dec 16, 2009 at 12:39 PM, Li Bing <lbl...@gmail.com> wrote: > Dear Didy, > > My solution is as follows. > > 1) Only one thread can index (write); > > 2) Multiple threads can retrieve (read). > > Now the retrieval performance is fine. Do you think my solution is > correct? Any better solutions? > > Best, > LB > > On Thu, Dec 17, 2009 at 1:36 AM, Digy <digyd...@gmail.com> wrote: > > I remember that this was discussed in thread "Exceptions When Indexing > with > > Multi-threading" Oct 2009 > > DIGY > > > > -----Original Message----- > > From: Li Bing [mailto:lbl...@gmail.com] > > Sent: Wednesday, December 16, 2009 11:46 AM > > To: lucene-net-user@incubator.apache.org > > Subject: What About the Performance When Retrieving at Indexing? > > > > Dear all, > > > > I just want to know what about the performance when retrieving while > > new data is being indexed if I do not manage the threads myself. It > > must be lower than retrieving an index which is not being updated, > > right? In my case, the retrieving can be processed after the indexing > > is done. It is really slow. Does Lucene solve the issue itself? > > > > Thanks again! > > LB > > > > ---------- Forwarded message ---------- > > From: Li Bing <lbl...@gmail.com> > > Date: Wed, Dec 16, 2009 at 5:26 PM > > Subject: How To Append and Synchronize Lucene Index? > > To: lucene-net-user@incubator.apache.org > > > > > > Hi, all, > > > > I am using Lucene.NET to index data. Meanwhile, more new data is > > frequently added to the index. However, if so, the retrieval > > performance becomes much lower when new data is being added and > > indexed because the indexing (writing) thread must synchronize the > > retrieving (reading) threads. > > > > I use .NET locks to manage the threads myself. Is it necessary? Or > > Lucene can handle the synchronize itself? > > > > I would like to know if it is possible for me to append one portion of > > Lucene indexes to another one. If so, it is easy to replicate Lucene > > indexes and the writing and reading do not affect each other. Or any > > other solutions in this case? > > > > Thanks so much! > > LB > > > > >
