[
https://issues.apache.org/jira/browse/DERBY-5752?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Knut Anders Hatlen updated DERBY-5752:
--------------------------------------
Attachment: buffsize.diff
I experimented with the attached buffsize.diff patch (not for commit), which
simply sets the buffer size to LOBStreamControl.DEFAULT_MAX_BUF_SIZE (4K)
regardless of the size of the LOB. With that patch, the heap requirement for
TriggerTest, which works on some large LOBs that get materialized, went down
from 110MB to 85MB. (I ran this experiment on sources that had been patched for
DERBY-5751, which already had reduced the heap requirements from 140MB to
110MB.) suites.All also passed with these changes.
This is not the approach I'm planning to take in the final patch. I intend to
make the buffer size dynamic based on the LOB size, as it is today, but have a
maximum size of 32KB.
Since the constructor in question is only used when the underlying LOB value is
materialized before the EmbedBlob is instantiated, and LOBs larger than 32KB
are typically not materialized at that point, I think the suggested approach
would only affect the cases where
- the LOB is used in a trigger
- getBytes() has been called on the column before getBlob() (this case was in
fact something we considered disallowing in DERBY-5489)
In those two cases, if the LOB is larger than 32KB, the LOBStreamControl
instance will overflow to temporary files instead of buffering the entire LOB
in memory, if the suggested approach is implemented.
> LOBStreamControl should materialize less aggressively
> -----------------------------------------------------
>
> Key: DERBY-5752
> URL: https://issues.apache.org/jira/browse/DERBY-5752
> Project: Derby
> Issue Type: Improvement
> Components: JDBC
> Affects Versions: 10.9.0.0
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Attachments: buffsize.diff
>
>
> The constructor LOBStreamControl(EmbedConnection, byte[]) always makes the
> buffer size equal to the LOB size, effectively creating an extra, fully
> materialized copy of the LOB in memory.
> I think the assumption here is that a LOB that's already materialized is a
> small one. That is, LOBs that are smaller than 32 KB and fit in a single page
> are typically materialized when read from store. However, we sometimes
> materialize LOBs that are a lot bigger than 32 KB. For example, triggers that
> access LOBs may materialize them regardless of size (see comment in
> DMLWriteResultSet's constructor for details). For these large LOBs, it sounds
> unreasonable to allocate a buffer of the same size as the LOB itself.
> I'd suggest that we change the constructor so that it never allocates a
> buffer larger than 32KB. That would mean that the behaviour is preserved for
> all LOBs fetched directly from store (only LOBs that don't fit in a single
> page will cause temporary files to be created), whereas we'll prevent large
> LOBs accessed by triggers from being duplicated in memory by overflowing to
> temporary files.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
