Speeding-up Fault Injection Experiments with Dynamic Code Injection

摘要

One of the main advantages of the software implemented fault injection methodology is a very high level of controllability of an injected fault and high observability of its propagation. However, in practice, the time cost of controllability of the application under tests and injected faults might be very high. This relates mainly to the problem of the proper fault injection time instant identification. The paper presents a new technique based on the dynamic code injection at the test runtime called Code Cave. The injected code, executed in the context of application under tests, traps the application at the precisely defined time instant. The comparative experiments with old-fashioned implementation show the gained speed-up reducing the experiment duration up to 50%.