db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andreas Korneliussen (JIRA)" <derby-...@db.apache.org>
Subject [jira] Updated: (DERBY-1559) when receiving a single EXTDTA object representing a BLOB, the server do not need to read it into memory before inserting it into the DB
Date Wed, 23 Aug 2006 14:58:15 GMT
     [ http://issues.apache.org/jira/browse/DERBY-1559?page=all ]

Andreas Korneliussen updated DERBY-1559:
----------------------------------------

    Attachment: DERBY-1559v3.diff

Attached is an updated patch were the conflicts from DERBY-1610 has been resolved.

Additionally, to avoid the failures in jdbcapi/parameterMapping, I have modified the code
so that it does only use setBinaryStream if it is a BLOB columns, otherwise it continues to
use setBytes.

I have also done some memory profiling, were the code inserts one BLOB of size 64 MB streamed
from the network client. 

The results are:
Without any changes: max memory usage 350 MB

With a special patch were setBytes was replaced by setBinaryStream: 176 MB

With this patch: 40 MB.


> when receiving a single EXTDTA object representing a BLOB, the server do not need to
read it into memory before inserting it into the DB
> ----------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-1559
>                 URL: http://issues.apache.org/jira/browse/DERBY-1559
>             Project: Derby
>          Issue Type: Sub-task
>          Components: Network Server
>    Affects Versions: 10.2.1.0, 10.3.0.0, 10.2.2.0
>            Reporter: Andreas Korneliussen
>         Assigned To: Andreas Korneliussen
>         Attachments: DERBY-1559.diff, DERBY-1559.stat, DERBY-1559v2.diff, DERBY-1559v3.diff,
serverMemoryUsage.xls
>
>
> When streaming a BLOB from the Network Client to the Network Server, the Network server
currently read all the data from the stream and put it into a byte array.
> The blob data is then inserted into the DB by using
> PreparedStatement.setBytes(..)
> and later
> PreparedStatement.execute()
> To avoid OutOfMemoryError if the size of the Blob is > than total memory in the VM,
we could make the network server create a stream which reads data when doing PreparedStatement.execute().
 The DB will then stream the BLOB data directly from the network inputstream into the disk.
> I intent to make a patch which does this if there is only one EXTDTA object (BLOB) sent
from the client in the statement, as it will simplify the implementation. Later this can be
improved  further to include CLOBs, and possibly to include the cases where there are multiple
EXTDTA objects.
> --
> CLOBs are more complex, as there need to be some character encoding. This can be achieved
by using a InputStreamReader,  and use PreparedStatement.setCharacterStream(..). However the
size of the stream is not necessarily the same as the size of the raw binary data, and to
do this for CLOBs, I would need the embedded prepared statements to support the new setCharacterStream()
overloads in JDBC4 (which do not include a length atribute)
> --
> Multiple EXTDATA objects are also more complex, since one would need to have fully read
the previous object ,before it is possible to read the next.
> --

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message