lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Koji Sekiguchi" <koji.sekigu...@m4.dion.ne.jp>
Subject Re: What is MMapDirectory?
Date Tue, 11 Oct 2005 07:25:17 GMT
Paul,

Thank you very much for your explanation.

> However, in case you have different experience, we'd like to know.

I don't have it. I'm just curious.

Thank you again,

Koji


----- Original Message ----- 
From: "Paul Elschot" <paul.elschot@xs4all.nl>
To: <java-user@lucene.apache.org>
Sent: Tuesday, October 11, 2005 4:16 PM
Subject: Re: What is MMapDirectory?


> Koji,
>
> On Sunday 09 October 2005 14:12, Koji Sekiguchi wrote:
>> Hello,
>>
>> What is MMapDirectory?
>>
>> I've searched mailing list archive, but cannot find it.
>> I could find the following explanation at Lucene 1.9 CHANGES.txt:
>>
>>  8. Add MMapDirectory, which uses nio to mmap input files.  This is
>>     still somewhat slower than FSDirectory.  However it uses less
>>     memory per query term, since a new buffer is not allocated per
>>     term, which may help applications which use, e.g., wildcard
>>     queries.  It may also someday be faster. (cutting & Paul Elschot)
>>
>> And I saw bug 31508:
>>
>> [PATCH] MMapDirectory for huge index files
>> http://issues.apache.org/bugzilla/show_bug.cgi?id=31508
>
> Bugzilla has been replaced by JIRA recently, this is now here:
> http://issues.apache.org/jira/browse/LUCENE-283
>
>>
>> But I'm curious to know about MMapDirecotry in more detail about
>> the following point:
>>
>> 1. Were there any troubles in FSDirectory with huge index files?
>
> There were some troubles with huge index files, but not in FSDirectory.
> FSDirectory is still faster than MMapDirectory, but that might change
> in the future.
>
>> 2. Is MMapDirectory used for only searching? (not indexing?)
>
> Indeed so.
>
>> 3. Why is MMapDirectory somewhat slower than FSDirectory?
>> 4. Why does MMapDirectory use less memory while using mapped file (mmap)?
>
> Both depend on the underlying implementation of buffering in the JVM
> and in the operating system. In my (few) attempts so far,  memory
> mapped I/O does not currently help to improve performance of Lucene.
> However, in case you have different experience, we'd like to know.
>
> Regards,
> Paul Elschot
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
> 



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


Mime
View raw message