Browse Prior Art Database

Method to minimize page contention for table insert operations in a shared disk database

IP.com Disclosure Number: IPCOM000206069D
Publication Date: 2011-Apr-13
Document File: 2 page(s) / 32K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a process providing capabilities to minimize page contention in a shared disk database system for insert operations.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.

Page 01 of 2

Method to minimize page contention for table insert operations in a shared disk database

Disclosed is a process providing capabilities to minimize page contention in a shared disk database system for insert operations.

In a database configuration where multiple systems (members) share objects through the use of a common storage device, contention on pages within database objects can negatively impact the performance of the system. Contention manifests as members waiting to be granted access to a page while the page continually moves from one member to another. Page contention can result from many routine operations in the database, and can also be a result of the layout of the database object.

Not all cases of page contention can be eliminated, but certain operations can be changed to use algorithms that help to reduce or largely avoid this type of contention. Using the disclosed process provides a capability to minimize contention on table insert operations for a shared disk database.

The disclosed process refers to a 'cluster accelerator', which is a system that is part of the database cluster, and through which operations such as the acquisition of a page is coordinated. Typically one of two approaches is used for inserting data into a table. Either an inserter searches through existing pages of the table for space or the inserter creates a new page associated with the object enabling inserted rows to be added. Without changes in a shared disk database all systems attempt to find space in, and insert data into the same set of pages, causing the pages to become hotly contended. The disclosed process eliminates page contention for concurrent inserts into a table from multiple members. Prior solutions typically implement optimizations for inserts on clustered systems, but require the optimizations to be explicitly activated, enabled by the user or serialize access to large portions of the table. The disclosed process enables optimizations for all tables at a fine level of page granularity.

When inserting data into a table using existing pages, an inserter searches through existing pages attempting to find room for the new data. In a shared disk database inserters running on different systems should not search through, or attempt to insert into the same range of pages. One member should also not search a range of pages already searched by another member. To solve these problems, when an inserter running on a particular system wants to insert data into a table, the inserter requests a range of pages to search in the object from the cluster accelerator. Responsive to searching this range of pages no space is found, a subsequent request to the cluster accelerator is made for another range of pages. When space is found, the insert proceeds, and subsequent inserts...