Mailing List Archive

Hello,

There is currenly no support for remote index searching, although you
will find some notes about it in Lucene (-user or -dev) mailing list
archives. I think the keyword to use in the search could be
'distributed' in the subject line - see http://nagoya.apache.org/.

I had a similar setup at the previous employer. I simply distributed
the index directory and its content to remote servers running Resin
using sdist/rsync. It worked pretty well, as the IndexSearcher
instance was able to tell when the index was updated and do what it
needed to do in that scenario.

Otis


> I have a web application deployed in a cluster of resin servers.
> Since
> Lucene works on a file system, do I need to mount a central drive on
> the
> machines with resin?
>
> Is there a way I can install Lucene in a central PC, and all clients
> from
> resin call a remote method (eg RMI) to perform a search?
>
> Any feedback would be appreciated.
>
> Regards
> Keith
>
>
> --
> To unsubscribe, e-mail:
> <mailto:lucene-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail:
> <mailto:lucene-user-help@jakarta.apache.org>
>


__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com

--
To unsubscribe, e-mail: <mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:lucene-user-help@jakarta.apache.org>

I found a SQLDirectory class from the mailing list. This will solve the
distributed problem, but it is reliable and fast?

Anyone has any feedback on using RDBMS instead of the file system?

Keith

-----Original Message-----
From: Otis Gospodnetic [mailto:otis_gospodnetic@yahoo.com]
Sent: Monday, 17 June 2002 10:03 a.m.
To: Lucene Users List
Subject: Re: multiple clients?


Hello,

There is currenly no support for remote index searching, although you
will find some notes about it in Lucene (-user or -dev) mailing list
archives. I think the keyword to use in the search could be
'distributed' in the subject line - see http://nagoya.apache.org/.

I had a similar setup at the previous employer. I simply distributed
the index directory and its content to remote servers running Resin
using sdist/rsync. It worked pretty well, as the IndexSearcher
instance was able to tell when the index was updated and do what it
needed to do in that scenario.

Otis


> I have a web application deployed in a cluster of resin servers.
> Since
> Lucene works on a file system, do I need to mount a central drive on
> the
> machines with resin?
>
> Is there a way I can install Lucene in a central PC, and all clients
> from
> resin call a remote method (eg RMI) to perform a search?
>
> Any feedback would be appreciated.
>
> Regards
> Keith
>
>
> --
> To unsubscribe, e-mail:
> <mailto:lucene-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail:
> <mailto:lucene-user-help@jakarta.apache.org>
>


__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com

--
To unsubscribe, e-mail:
<mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail:
<mailto:lucene-user-help@jakarta.apache.org>


--
To unsubscribe, e-mail: <mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:lucene-user-help@jakarta.apache.org>

I thought of using that, too, when faced with your issue.
I contacted the author with the same questions that you have, and the
answers were not encouraging. Try using my first name and SQLDirectory
as keywords for the list archive search and you may be able find
another message from me with more details about the author's response.

Otis

--- Keith Chew <keith.chew@jungledrum.co.nz> wrote:
>
> I found a SQLDirectory class from the mailing list. This will solve
> the
> distributed problem, but it is reliable and fast?
>
> Anyone has any feedback on using RDBMS instead of the file system?
>
> Keith
>
> -----Original Message-----
> From: Otis Gospodnetic [mailto:otis_gospodnetic@yahoo.com]
> Sent: Monday, 17 June 2002 10:03 a.m.
> To: Lucene Users List
> Subject: Re: multiple clients?
>
>
> Hello,
>
> There is currenly no support for remote index searching, although you
> will find some notes about it in Lucene (-user or -dev) mailing list
> archives. I think the keyword to use in the search could be
> 'distributed' in the subject line - see http://nagoya.apache.org/.
>
> I had a similar setup at the previous employer. I simply distributed
> the index directory and its content to remote servers running Resin
> using sdist/rsync. It worked pretty well, as the IndexSearcher
> instance was able to tell when the index was updated and do what it
> needed to do in that scenario.
>
> Otis
>
>
> > I have a web application deployed in a cluster of resin servers.
> > Since
> > Lucene works on a file system, do I need to mount a central drive
> on
> > the
> > machines with resin?
> >
> > Is there a way I can install Lucene in a central PC, and all
> clients
> > from
> > resin call a remote method (eg RMI) to perform a search?
> >
> > Any feedback would be appreciated.
> >
> > Regards
> > Keith
> >
> >
> > --
> > To unsubscribe, e-mail:
> > <mailto:lucene-user-unsubscribe@jakarta.apache.org>
> > For additional commands, e-mail:
> > <mailto:lucene-user-help@jakarta.apache.org>
> >
>
>
> __________________________________________________
> Do You Yahoo!?
> Yahoo! - Official partner of 2002 FIFA World Cup
> http://fifaworldcup.yahoo.com
>
> --
> To unsubscribe, e-mail:
> <mailto:lucene-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail:
> <mailto:lucene-user-help@jakarta.apache.org>
>
>
> --
> To unsubscribe, e-mail:
> <mailto:lucene-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail:
> <mailto:lucene-user-help@jakarta.apache.org>
>


__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com

--
To unsubscribe, e-mail: <mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:lucene-user-help@jakarta.apache.org>

Hi,

> I found a SQLDirectory class from the mailing list. This will solve the
> distributed problem, but it is reliable and fast?
>
> Anyone has any feedback on using RDBMS instead of the file system?

We made comparison of FSDirectory vs. SQLDirectory, and found that version
that was available in mailing list is very slow. Phrase search in 5,000
documents is performed in 601 ms in FSDirectory and 133813 ms in
SQLDirectory connected to PostgreSQL. But I'm pretty sure one can improve
SQLDirectory, for example, 1k block size is too small, because time to
execute query is greater than to transfer data over the wire.

Our next try will be JDataStore from Borland, it seems very promissing (in
my DB performance tests it outperformed PostgreSQL, SAP DB and InterBase, as
well as all-in-java DBMS-es). But JDataStore is not free.

Best regards,
Roman Rokytskyy


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com


--
To unsubscribe, e-mail: <mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:lucene-user-help@jakarta.apache.org>

Hi Roman

Thank you for your feedback.

I will experiment with FSDirectory first, and get the code up and running.

Keith



-----Original Message-----
From: Roman Rokytskyy [mailto:rrokytskyy@yahoo.co.uk]
Sent: Monday, 17 June 2002 8:40 p.m.
To: Lucene Users List
Subject: Re: multiple clients?


Hi,

> I found a SQLDirectory class from the mailing list. This will solve the
> distributed problem, but it is reliable and fast?
>
> Anyone has any feedback on using RDBMS instead of the file system?

We made comparison of FSDirectory vs. SQLDirectory, and found that version
that was available in mailing list is very slow. Phrase search in 5,000
documents is performed in 601 ms in FSDirectory and 133813 ms in
SQLDirectory connected to PostgreSQL. But I'm pretty sure one can improve
SQLDirectory, for example, 1k block size is too small, because time to
execute query is greater than to transfer data over the wire.

Our next try will be JDataStore from Borland, it seems very promissing (in
my DB performance tests it outperformed PostgreSQL, SAP DB and InterBase, as
well as all-in-java DBMS-es). But JDataStore is not free.

Best regards,
Roman Rokytskyy


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com


--
To unsubscribe, e-mail:
<mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail:
<mailto:lucene-user-help@jakarta.apache.org>


--
To unsubscribe, e-mail: <mailto:lucene-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:lucene-user-help@jakarta.apache.org>