A critical section of code makes a set of disk modifications that must be performed as a single unit; either all the modifications must occur, or none can occur.
A thread that is in a critical section is holding shared-memory resources. Within the space of the critical section, the database server cannot determine which shared-memory resources should be released and which changes should be undone to return all data to a consistent point. Therefore, if a virtual processor is terminated while a thread is in a critical section, the database server takes the following two steps to ensure that all data is returned to the last known point of consistency:
Fast recovery is the procedure that the database server uses to restore the physical and logical consistency of data quickly, up to and including the last record in the logical log. For a description of fast recovery, refer to Fast Recovery.
Home | [ Top of Page | Previous Page | Next Page | Contents | Index ]