Information Technology Reference
In-Depth Information
A Causality Analysis Framework
for Component-Based Real-Time Systems
Shaohui Wang
1
, Anaheed Ayoub
1
, BaekGyu Kim
1
, Gregor Gossler
2
,
Oleg Sokolsky
1
,andInsupLee
1
1
Department of Computer and Information Science
University of Pennsylvania
{
shaohui,anaheed,baekgyu
}
@seas.upenn.edu,
{
sokolsky,lee
}
@cis.upenn.edu
2
INRIA Grenoble - Rhone-Alpes, France
gregor.goessler@inria.fr
Abstract.
We propose an approach to enhance the fault diagnosis in
black-box component-based systems, in which only events on component
interfaces are observable, and assume that causal dependencies between
component interface events within components are not known. For such
systems, we describe a causality analysis framework that helps us estab-
lish the causal relationship between component failures and system fail-
ures, given an observed system execution trace. The analysis is based on
a formalization of counterfactual reasoning, and applicable to real-time
systems. We illustrate the analysis with a case study from the medical
device domain.
1 Introduction
Component-based design in systems engineering enables independent develop-
ment of system components as well as their incremental construction and modi-
fication. The complexity of systems that are built with component-based design
renders it dicult to determine the culprit components of the system that are
responsible for the discovered system failure on a given system execution. We in
this paper aim to present a formal framework for the analysis of the causal rela-
tion between the faulty components and an observed system failure on a given
system execution.
While this problem is common to all safety-critical domains, our immediate
motivation comes from the domain of medical devices. In the United States,
the Food and Drug Administration (FDA) is responsible for assessing safety of
medical devices and regulating their use in health care. When a system failure
that harms a patient, known as an
adverse event
occurs, the hospital is required
to report it to the FDA-maintained database [9]. Diagnosis of the root cause
is crucial for the subsequent recovery and follow-up prevention measures. Such
diagnosis requires recording of system executions leading to the failure, as well
as methods for the ecient analysis of the recorded system trace.
Research is supported in part by the National Science Foundation grants CNS-
0930647 and CNS-1035715, and NSF/FDA SiR grant CNS-1042829.
