Mailing-List: contact lucene-dev-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Lucene Developers List" <lucene-dev@jakarta.apache.org>
From: "Aviran" <amordo@infosciences.com>
To: "'Lucene Developers List'" <lucene-dev@jakarta.apache.org>
Subject: RE: FW: Lucene Search has poor cpu utilization on a 4-CPU machine
Date: Thu, 15 Jul 2004 09:34:55 -0400
Message-ID: <002201c46a70$87074ae0$6a00a8c0@Aviran>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
In-Reply-To: <40F59B2F.4050402@apache.org>
Importance: Normal


> The second one is on
> org.apache.lucene.index.SegmentReader.norms(SegmentReader.java:318)=20
> which is a synchronized method thus causing locks. I guess the=20
> synchronization is done for a good reason, but you probably know the=20
> answer better then me.

I'm surprised this is showing up.  Can you tell more about the size of=20
your index and the nature of your queries?  If you're, e.g., doing lots=20
of range or wildcard queries, then I can maybe see this showing up a=20
little.  What is your benchmark like?

Are you "warming the cache" when you're performing these benchmarks?  In =

other words, are you first sending a few queries at a low rate before=20
you start slamming it with high traffic?  If you're not, and/or you have =

a lot of fields, or you re-open searchers a lot, then this could show up =

too.

Doug


My test index is pretty small size, about 250 documents and about 24 =
fields
in each document.=20
The test is done by starting 10 threads that repeat simple one word =
query
(each thread query on a different word). Neither range nor wildcard =
query is
done.
I let the test run for about a minute and then I do a full thread dump =
to
see the stack trace.
I use a single searcher which never gets closed.

Aviran


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