Good. Last question. Does Nutch guaranty that it will create new index for
each separate segment? May it happen that it create one index for two
segments or vise-versa.

Alexander

2008/8/4 Andrzej Bialecki <[EMAIL PROTECTED]>

> Alexander Aristov wrote:
>
>> Hi
>>
>> Thank you for Katta
>>
>> But are there any built-in Nutch functionality which can do this stuff.
>> What
>> I am looking forward is to make distributed search as I am planning to
>> build
>> an index of quite big size and so it will be not possible to keep it on
>> one
>> server.
>>
>> What are best practices for doing this?
>>
>
> There is no built-in single tool in Nutch to do this. Common practice is to
> create indexes per segment (without merging them), and deploy pairs of
> segment plus its index to the search servers, and then doing the index
> merging there, on each search server. Whenever you add new segments or
> remove old ones, you perform a merge of the new set of active indexes on
> each search server.
>
> This way it's easy to phase out outdated segments and their indexes, and
> adding new segments, while still using a merged index on each search server
> for maximum performance.
>
> PS. it's possible to implement a low-level Lucene tool to split indexes,
> using FilteredIndexReader and IndexWriter.addIndexes(...). But it's not that
> relevant if you use the strategy that I explained above.
>
>
> --
> Best regards,
> Andrzej Bialecki     <><
>  ___. ___ ___ ___ _ _   __________________________________
> [__ || __|__/|__||\/|  Information Retrieval, Semantic Web
> ___|||__||  \|  ||  |  Embedded Unix, System Integration
> http://www.sigram.com  Contact: info at sigram dot com
>
>


-- 
Best Regards
Alexander Aristov

Reply via email to