On Tue, Jun 26, 2007 at 09:29:40AM +0200, Peter De Berdt wrote:
Btw, what exactly does ‘Optimized for Low to High Volume Web Traffic’
Ferret is fine for low traffic sites, but once traffic goes up,
things start to happen. I’m sure people on this list will know the
disastrous “segmentation fault” errors ferret is well known for. The
latest version of ferret has dealt with a lot of these bugs, so
things should be better.
Recent Ferret only segfaults or corrupts the index when multiple
processes write-access the index at the same time. This is a well-known
and discussed issue and one that can easily be solved by doing the
indexing with a single process, i.e. some backend server solution just
as solr is one, too.
Also, ferret has no way of splitting up your indexing to balance it
over another server.
Right, Ferret doesn’t have that feature. I didn’t miss it, yet.
All in all, I’ve switched from ferret to solr and although I’m not a
big java fan because of its memory consumption, solr and the real
lucene behind it blow away ferret both in terms of speed and
Acts_as_ferret brings an indexing server that will handle all indexing
in production environments, so you’ll see no segfaults or other problems
at all. Plus you get features like index versioning and background
For the speed, Solr and acts_as_ferret/DRb seem to be pretty up to par,
with a slight advantage for Ferret according to this benchmark (not done
by me, but by a user of aaf):
I’m not at all into search engine flamewars, but please hold to the
facts and don’t post any unproven generalized speed estimates.
Regarding the stability you might like to have a look at projects like
omdb.org which runs Ferret for lots of live searches and the main site
search. it has a background indexer that is busy reindexing stuff as
soon as something on the page is changed - it runs without a single
hickup for months now. The project’s code is MIT licensed, so you can
even have a look at how it’s done under the hood.
http://www.jkraemer.net/ - Blog
http://www.omdb.org/ - The new free film database