Обсуждение: Text Search vs MYSQL vs Lucene

What would be performance of pgSQL text search vs MySQL vs Lucene (flat
file) for a 2 terabyte db?
thanks for any comments.
.V
--
Please post on Rich Internet Applications User Interface (RiA/SoA)
<http://www.portalvu.com>

On Thursday 09 Sep 2004 6:26 pm, Vic Cekvenich wrote:
> What would be performance of pgSQL text search vs MySQL vs Lucene (flat
> file) for a 2 terabyte db?

Well, it depends upon lot of factors. There are few questions to be asked
here..
- What is your hardware and OS configuration?
- What type of data you are dealing with? Mostly static or frequently updated?
- What type of query you are doing. Aggregates or table scan or selective
retreival etc.

Unfortunately there is no one good answer. If you could provide details, it
would help a lot..

 Shridhar

On Thu, Sep 09, 2004 at 07:56:20AM -0500, Vic Cekvenich wrote:

> What would be performance of pgSQL text search vs MySQL vs Lucene (flat
> file) for a 2 terabyte db?
> thanks for any comments.

My experience with tsearch2 has been that indexing even moderately
large chunks of data is too slow to be feasible. Moderately large
meaning tens of megabytes.

Your milage might well vary, but I wouldn't rely on postgresql full
text search of that much data being functional, let alone fast enough
to be useful. Test before making any decisions.

If it's a static or moderately static text corpus you're probably
better using a traditional FTS system anyway (tsearch2 has two
advantages - tight integration with pgsql and good support for
incremental indexing).

Two terabytes is a lot of data. I'd suggest you do some research on
FTS algorithms rather than just picking one of the off-the-shelf FTS
systems without understanding what they actually do. "Managing
Gigabytes" ISBN 1-55860-570-3 covers some approaches.

Cheers,
  Steve

It be at least dual opteron 64 w 4 gigs of ram runing fedora with a huge
raid striped drives as single volume.
A similar system and types of querries would be this:
http://marc.theaimsgroup.com

So I guess a table scan.


.V

Shridhar Daithankar wrote:

>On Thursday 09 Sep 2004 6:26 pm, Vic Cekvenich wrote:
>
>
>>What would be performance of pgSQL text search vs MySQL vs Lucene (flat
>>file) for a 2 terabyte db?
>>
>>
>
>Well, it depends upon lot of factors. There are few questions to be asked
>here..
>- What is your hardware and OS configuration?
>- What type of data you are dealing with? Mostly static or frequently updated?
>- What type of query you are doing. Aggregates or table scan or selective
>retreival etc.
>
>Unfortunately there is no one good answer. If you could provide details, it
>would help a lot..
>
> Shridhar
>
>---------------------------(end of broadcast)---------------------------
>TIP 2: you can get off all lists at once with the unregister command
>    (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)
>
>
>
>


--
Please post on Rich Internet Applications User Interface (RiA/SoA)
<http://www.portalvu.com>

Steve Atkins wrote:
>>What would be performance of pgSQL text search vs MySQL vs Lucene (flat
>>file) for a 2 terabyte db?
>>thanks for any comments.
>
> My experience with tsearch2 has been that indexing even moderately
> large chunks of data is too slow to be feasible. Moderately large
> meaning tens of megabytes.

My experience with MySQL's full text search as well as the various
MySQL-based text indexing programs (forgot the names, it's been a while)
for some 10-20GB of mail archives has been pretty disappointing too. My
biggest gripe is with the indexing speed. It literally takes days to
index less than a million documents.

I ended up using Swish++. Microsoft's CHM compiler also has pretty
amazing indexing speed (though it crashes quite often when encountering
bad HTML).

--
dave

I'd say indexing of 2 TB of data would be a very costly even for
standalone solution ( no relational database ).
Ideal solution would be to have tsearch2 for current documents and
standalone solution for archive documents. If these solutions share
common parsers,dictionaries and ranking schemes it would be easy to
combine results from two queries. We have prototype for standalone
solution - it's based on OpenFTS, which is already tsearch2 compatible.


    Oleg
On Thu, 9 Sep 2004, Steve Atkins wrote:

> On Thu, Sep 09, 2004 at 07:56:20AM -0500, Vic Cekvenich wrote:
>
> > What would be performance of pgSQL text search vs MySQL vs Lucene (flat
> > file) for a 2 terabyte db?
> > thanks for any comments.
>
> My experience with tsearch2 has been that indexing even moderately
> large chunks of data is too slow to be feasible. Moderately large
> meaning tens of megabytes.
>
> Your milage might well vary, but I wouldn't rely on postgresql full
> text search of that much data being functional, let alone fast enough
> to be useful. Test before making any decisions.
>
> If it's a static or moderately static text corpus you're probably
> better using a traditional FTS system anyway (tsearch2 has two
> advantages - tight integration with pgsql and good support for
> incremental indexing).
>
> Two terabytes is a lot of data. I'd suggest you do some research on
> FTS algorithms rather than just picking one of the off-the-shelf FTS
> systems without understanding what they actually do. "Managing
> Gigabytes" ISBN 1-55860-570-3 covers some approaches.
>
> Cheers,
>   Steve
>
> ---------------------------(end of broadcast)---------------------------
> TIP 5: Have you checked our extensive FAQ?
>
>                http://www.postgresql.org/docs/faqs/FAQ.html
>

    Regards,
        Oleg
_____________________________________________________________
Oleg Bartunov, sci.researcher, hostmaster of AstroNet,
Sternberg Astronomical Institute, Moscow University (Russia)
Internet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/
phone: +007(095)939-16-83, +007(095)939-23-83