| 摘要 |
An improved method to dramatically reduce the time required to reorganize DB2 tablespaces and index files by not utilizing conventional sort techniques. Viewing access is allowed during the reorganization process by setting the files to read only status. The process is basically non-destructive, allowing a prompt return to the original state, and is checkpointed, allowing restarting at selected intervals. Briefly, the original table and indices are considered as A files and read into memory. New row IDs or RIDs are developed using a non-sorting technique so that the proper order of the data is developed. After the new RIDs have been developed, both the clustering index and the row data are read out of memory and written to a new table and clustering index files in the proper order as B files. All of the table files are then stopped to allow exclusive access. Next, a series of AMS statements are built to do the renaming operations. Specifically, a series of statements are built to rename all of the A files to X files, to rename all B files to A files and then to delete all of the X files. Then any remaining non-clustering indices are reorganized using non-sorting techniques and renamed in a similar fashion.
|
