Operating System Support to Detect Application Hangs

Abstract

On-line failure detection is an essential means to control and assess the dependability of complex and critical software systems. In such context, effective detection strategies are required, in order to minimize the possibility of catastrophic consequences. This objective is however difficult to achieve in complex systems, especially due to the several sources of non-determinism (e.g., multi-threading and distributed interaction) which may lead to software hangs, i.e., the system is active but no longer capable of delivering its services. The paper proposes a detection approach to uncover application hangs. It exploits multiple indirect data gathered at the operating system level to monitor the system and to trigger alarms if the observed behavior deviates from the expected one. By means of fault injection experiments conducted on a research prototype, it is shown how the combination of several operating system monitors actually leads to an high quality of detection, at an acceptable overhead.

A monitoring and testing framework for critical off-the-shelf applications and services

“…Finally, the toolset will include ready to use tools that use some more specific algorithms targeting fault detection, proposed by research community. Examples of these works are static threshold analysis [9] and statistical analysis algorithms for on-line fault detection [8].…”

“…Statistical analysis algorithms have been used in the past for on-line fault detection [8]. This technique overcomes some of the limitations of static threshold analysis, that for instance in [9] monitoring techniques are used to detect application hangs. Works towards certifying OTS components are also not new.…”

Thread Homeostasis - Real-Time Anomalous Behavior Detection Using Short Sequences of Messages for Safety-Critical Software

“…We would rather have a system revert to its design or fail-safe state if it is performing the incorrect function or misbehaving than have a lively system that never hangs but executes erroneously. Furthermore, in some cases, the hang detection algorithms require a pre-determined threshold that is tuned during the training phase such as the work by Carrozza et al [15]. This is not very well suited for systems with highly dynamic environments and might require adaptive threshold techniques like the one presented by Bovenzi et al [13].…”

“…These methods are often concerned with the liveness property of a system versus its correctness. A hang condition can be either a result of infinite loops, also known as active hangs, or a result of permanent or extended wait conditions, known as passive hangs [15]. Irrera et al [31] argue that critical systems are too complex and have highly non-deterministic behavior that renders online fault detection for safetycritical systems too difficult of a task.…”