Browse Prior Art Database

Pre-Failure Memory Error Determination during Run-Time

IP.com Disclosure Number: IPCOM000110808D
Original Publication Date: 1994-Jan-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 2 page(s) / 28K

Publishing Venue

IBM

Related People

Haselhorst, KH: AUTHOR [+3]

Abstract

This article describes a method for dynamically determining if small DRAM errors slow down the system and to determine how to repair them with steering. This is accomplished before memory failure, during normal run-time processing.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 100% of the total text.

Pre-Failure Memory Error Determination during Run-Time

      This article describes a method for dynamically determining if
small DRAM errors slow down the system and to determine how to repair
them with steering.  This is accomplished before memory failure,
during normal run-time processing.

      As each normal fetch is processed by the main store card, data
about each error is tracked.  The data kept is:

o   the  total number of errors,

o   the number of of errors for each bit in each ECC word, and

o   the number of errors for each block.

      The total number of errors is used along with a programmable
timer to determine if errors are slowing down the system.  If a
threshold of errors occurs before the timer expires, the error is
affecting system performance and must be handled.

      The other data kept is used to determine how to fix the
problem.  The actual size of the physical error is not measured;
however, as the system addressing pattern is used to find the errors,
the impact to the system as a whole is automatically taken into
account (Figure).