db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mike Matrigali (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DERBY-6809) Java 1.8 feature use
Date Wed, 20 May 2015 14:55:00 GMT

    [ https://issues.apache.org/jira/browse/DERBY-6809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14552425#comment-14552425
] 

Mike Matrigali commented on DERBY-6809:
---------------------------------------

i would add that you should make sure to not order by any key/key set that already has an
index as Derby will try it's
best to eliminate sorts by using the index.  For a test case easily done just by not having
any indexes.  The sample size
seems small to me but depends on data length.  I would suggest designing your test case to
be able to generate N rows
of data and then pick whatever order of magnitude number of rows takes seconds to run.  I
find that amount of work 
significant enough to usually give good results, but not too long so that you can do iterative
debugging easily.  Later
longer runs may be interesting.

There are 2 paths through the sorter.  One tries to do everything in memory, the other splits
the problem up into groups 
and writes/reads merge runs from/to disk.  The threshold is based on optimizer estimate of
amount of data.  Off hand I do
not remember the default and there is some knob that can be used to bump it up.  The default
is likely too small for most
applications that have huge amounts of memory.

> Java 1.8 feature use
> --------------------
>
>                 Key: DERBY-6809
>                 URL: https://issues.apache.org/jira/browse/DERBY-6809
>             Project: Derby
>          Issue Type: Improvement
>          Components: Network Server
>    Affects Versions: 11.0.0.0
>            Reporter: sagar
>
> Suggestion ...
> Is it possible to auto modify the existing source code using tools like Netbeans, and
take advantage of the new features in JDK 1.8 for better multiuser performance and better
utilization of current day multicore processors?
> Plainly put, can we have from 11.0 onwards a version of derby which takes advantage of
the advancements and new features in java 1.8 ... 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message