Archives of the TeradataForum

Message Posted: Fri, 22 Oct 2004 @ 19:37:54 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

John

Here is an excerpt from a white paper (not yet on www.teradata.com ).

Data warehouse environments depend on the integrity of the underlying physical database to provide a highly reliable and available data store. Erroneous answer sets or data unavailability that result from data corruption are extremely costly and disruptive to the enterprise.

While infrequent, data corruption can have many different forms and root causes. The I/O subsystem is comprised of many different components, and corruption can be caused by a problem anywhere on the path from the compute node to the disk or from the disk back to the compute node. Most data corruption that occurs during I/O transit is detected, and often corrected, by various parity, error checking, and correcting mechanisms in the hardware. If the corruption is detected, but cannot be corrected, the pending operation is failed and retried by the I/O subsystem software. Unfortunately, not all hardware devices in the I/O path have error detection and correction schemes in place, nor is the compute node and storage device firmware and software always flawless. Therefore, it is up to the Teradata Database to prevent, detect, correct, and recover from data corruption where possible.

Mark

	<Prev	Next>		<<First	<Prev	Next>	Last>>