Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Understanding Why Symptom Detectors Work by Studying Data-Only Application Values
 

Summary: 1
Understanding Why Symptom Detectors Work
by Studying Data-Only Application Values
Pradeep Ramachandran, Siva Kumar Sastry Hari, Sarita V. Adve, and Helia Naeimi
Department of Computer Science, University of Illinois at Urbana-Champaign
swat@cs.uiuc.edu
Intel Labs, Intel Corporation
helia.naeimi@intel.com
Abstract--Failures from unreliable hardware are posing a
serious threat to system reliability. Symptom detectors that
monitor anomalous software execution to detect such failures
are emerging as a viable low-cost detection scheme for future
systems. Since these detectors do not provide perfect resiliency
guarantees, they strive towards reducing instances when faults
escape detection and affect application output; such faults are
commonly referred to as Silent Data Corruptions (SDCs). Pre-
vious work on symptom detection has demonstrated low SDC
rates through empirical fault injection experiments.
This paper, for the first time, presents an intuitive reasoning
behind why symptom detectors achieve such low SDC rates. The

  

Source: Adve, Sarita - Department of Computer Science, University of Illinois at Urbana-Champaign

 

Collections: Computer Technologies and Information Sciences