-
Annamalai Gurusami authored
This is a follow up worklog for WL#8960. In WL#8960, the minimum change to a BLOB is a single LOB page. Even if only a few bytes are modified, minimum one LOB page will be modified. So, there is scope for improvement for small changes to the BLOBs. The solution provided by WL#8960 is a general solution suitable for all sizes of BLOBs and for all size of changes (from few bytes to even 1GB of changes). But it is less efficient for small changes done to the BLOBs. Also, in the case of WL#8960, the old versions of BLOBs are stored in the BLOB itself. Undo log format is not changed. To optimize small changes to the BLOBs, we plan to do the regular undo logging. For this the undo log format needs to be changed. When a BLOB is modified/updated, then we need to store the old and new portion of the BLOB in the undo log record. Currently there is a restriction that the undo log record must fit within an undo log page. We need to perform our operation within that constraint.
Annamalai Gurusami authoredThis is a follow up worklog for WL#8960. In WL#8960, the minimum change to a BLOB is a single LOB page. Even if only a few bytes are modified, minimum one LOB page will be modified. So, there is scope for improvement for small changes to the BLOBs. The solution provided by WL#8960 is a general solution suitable for all sizes of BLOBs and for all size of changes (from few bytes to even 1GB of changes). But it is less efficient for small changes done to the BLOBs. Also, in the case of WL#8960, the old versions of BLOBs are stored in the BLOB itself. Undo log format is not changed. To optimize small changes to the BLOBs, we plan to do the regular undo logging. For this the undo log format needs to be changed. When a BLOB is modified/updated, then we need to store the old and new portion of the BLOB in the undo log record. Currently there is a restriction that the undo log record must fit within an undo log page. We need to perform our operation within that constraint.
Loading