When the onutil repack data command is run against a raw table, the repack thread moves whole pages at a time from the tail of the table to the head, rather than moving individual rows one at a time. Depending on the size of the rows, this repack method has the potential to speed the process by 30 to 50 percent, because multiple rows are moved at a time, and because there is no need to record these operations in the logical log. This algorithm is faster, but it does not pack rows as tightly as the row-by-row method. The rows being moved are locked as the page they reside on is moved, which allows other work to be done on the table during the repack operation.
Note that this page-moving algorithm is efficient with raw tables that do not contain any rows that span more than one page. If a repack thread encounters a row that spans more than one page, the row-moving algorithm replaces the page-moving algorithm.
Enterprise Edition Home | Express Edition Home | [ Top of Page | Previous Page | Next Page | Contents | Index ]