lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Hostetter <hossman_luc...@fucit.org>
Subject Re: Web services for querying and return of results
Date Wed, 24 May 2006 18:10:52 GMT

: Solr sounds very interesting though - how do you maintain the cache?
: Are you storing filters?  And how do you persist these, via the session
: or using some kind of register?  I was considering designing something
: like this but felt that state management was something I wanted to avoid
: doing if possible, because it limits how many users we can service, if a
: lot of data is being kept round for caching.  Would be very interested
: in how you managed to do this, and still make Solr scalable.

this question would probably make more sense on the solr-user mailing
list, but the short answer is: No session state, just Caches of
Query=>DocSet when scoring isn't used, and Query => DocList (when scoring
is used).  The caches are maintained on a per searcher basis, and a single
searcher is used to serve all requests untill a "commit" is made, a new
searcher is opened and it's caches are "warmed" using the keys of the old
searchers cache before any user queries are sent to it.

(A DocSet is like a BitSet, but there is support for alternate compact
representations when the number of docs is small, a DocList is a "window"
into ordered results -- think a single page of paginated results)



-Hoss


---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org


Mime
View raw message