I am writing a Solr-powered OPAC right now and have not had any performance problems (either indexing or querying) using Solr for both data storage and search.  You can indicate in Solr whether you want particular data fields to be stored, indexed or both.  So I stick the entire MARC record in a Solr field but don't index on it.  
 
Just using Solr has proven to be much faster than doing the search in Solr and then retrieving full data from another database.  This also has the advantage of making it so there's only one thing you gotta keep in sync with the ILS.  The only data that my OPAC needs to talk to a SQL database for is item-level information, which changes too often to keep synced.
 
Searching Solr on the exact accession number of the record can be used for known item retrieval or for retrieving lists of items (such as patron booklists).  
 
--Casey


>>> [log in to unmask] 11/27/2006 1:00 PM >>>
On Nov 27, 2006, at 3:54 PM, Andrew Nagy wrote:

At UVa we have been burned several times by poor search performance
of XML native databases. In light of this, we're starting to look at
the database and the index as separate but cooperative pieces of the
application. That way you can use solr / lucene for search, faceted
browse, etc, and your XML database only for known item retrieval,
which it is generally able to do without performance issues. I'm
hopping up and down waiting for someone to take this approach with an
ILS, so please come and show us what you've got!

Bess