|
Activity Number:
|
189
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #302173 |
|
Title:
|
Predicting Neutron-induced Soft Errors
|
|
Author(s):
|
Sarah E. Michalak*+
|
|
Companies:
|
Los Alamos National Laboratory
|
|
Address:
|
, Los Alamos, NM, 87545,
|
|
Keywords:
|
inverse problem ; soft error ; supercomputer
|
|
Abstract:
|
A "soft error" is a nondestructive error that causes a change in the state of an electronic component. Bit flips in computer memory are a frequent example. Such bit flips may be caused by neutrons that are produced by cosmic rays hitting nuclei in the atmosphere. Neutron-induced soft errors are a major concern for systems that contain many susceptible parts (such as supercomputers) and for devices housed at very high altitudes where the neutron flux is quite high (such as satellites and airplanes). In such systems, soft errors can result in system crashes or incorrect information being passed to a program. Estimating a device's susceptibility to neutron-induced soft errors is crucial for assessing its feasibility and reliability in particular environments. This talk describes current methods for using the Los Alamos Neutron Science Center to estimate the impact of soft errors in the deployed setting and suggests a framework for providing more accurate predictions of the expected number of neutron-induced soft errors and an associated standard error. The framework involves a carefully designed set of experiments and the solution of an inverse problem.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2004 program |
