Disk read-write optimizations and data integrity in transaction systems using write-ahead logging

Abstract

Several disk read-write optimizations implemented in transaction systems and disk hardware to improve performance are discussed. These include writing multiple sectors written to disks out of sequence, avoiding initialization of pages on disk when a file is extended, not assessing individual pages during a mass delete operation, permitting a previously deallocated page to be reallocated without the need to read the deallocated version of the page from disk during reallocation, purging file pages from the buffer pool during a file erase operation and avoiding logging of bulk operations as index create. Following the discussion of various optimizations, a symbol method for guaranteeing detection of partial disk write of a page is presented for a system in which a page consists of multiple disk sectors and the recovery is based on write-ahead logging using a log sequence number on every page.