lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Marcus Falck" <>
Subject SV: Date first best practises
Date Sun, 21 May 2006 16:12:56 GMT
I preffer sorting. But I'm afraid that it won't be enough. 
How long time do you think it will take to rebuild the caches?
If I use a sort on the datefield and perform a query (with that sort) will it always rebuild
the whole cache or
just the cache for the actual hits?


Från: Erik Hatcher []
Skickat: sö 2006-05-21 17:45
Ämne: Re: Date first best practises

On May 21, 2006, at 11:31 AM, Marcus Falck wrote:
> I will use Lucene to index 200 million documents (doc size 2kb -> 
> 20 kb).
> With the following requirements:
> IndexSearcher needs to be created atleast every 5 minute.
> The ranking/scoring/sorting will need to reply the hits ordered by 
> date desc.
> Will the sorting be good enough on a machine with 4GB of RAM? Since 
> the cache will be destroyed every time I open a new IndexSercher i 
> need to know how long it will take to rebuild it (the index is 
> stored on a SAN).

My hunch is that sorting will be fine, but you'll want to warm up the 
caches before putting a new IndexSearcher into production use.  Solr 
does this by running a configurable set of queries against an index 
before swapping out the old IndexSearcher with a new warmed up one.

> Or do I have to change the scoring so I wont need to access the 
> date field?
> If i will have to change the scoring does anybody have any tips on 
> where to begin?

Solr's FunctionQuery (isn't it time we move this to the core of 
Lucene?! :) might be able to help you out in this situation, but it 
sounds like sorting is what you want to use, not scoring tweaks.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message