[
https://issues.apache.org/jira/browse/DERBY-5210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13029257#comment-13029257
]
Knut Anders Hatlen commented on DERBY-5210:
-------------------------------------------
I changed Utf8CcsidManager to use a CharsetEncoder to encode strings directly
into the ByteBuffer and reran the experiment (will post a patch on DERBY-5068
soon). With these changes, I saw reduced CPU usage compared to 10.8.1.2 (still
no observable change in throughput, though). So I think it's safe to assume
that although the 1a patch seen in isolation makes the client driver use
slightly more CPU, it enables us to add optimizations that outweigh the extra
cost. And, regardless of that, the extra cost seems to be so small that it
might be worthwhile to check in the patch in any case, just because it makes
the code simpler.
> Use java.nio.ByteBuffer in client.net.Request
> ---------------------------------------------
>
> Key: DERBY-5210
> URL: https://issues.apache.org/jira/browse/DERBY-5210
> Project: Derby
> Issue Type: Improvement
> Components: Network Client
> Affects Versions: 10.9.0.0
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Attachments: d5210-1a.diff
>
>
> We should see if we could use a java.nio.ByteBuffer instead of a byte array
> in org.apache.derby.client.net.Request, similar to what we did for DDMWriter
> in DERBY-2936. ByteBuffer provides some helper methods that allows us to
> simplify the code that puts multi-byte values into the buffer (like
> ByteBuffer.putShort(), putInt(), putLong()). It may also be a first step on
> the way to using a CharsetEncoder to encode strings without going via an
> intermediate throw-away byte array (see DERBY-5068 for details).
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
