Platform Software Reliability for Cloud Service Continuity - Challenges and Opportunities
Ning Luo, Yue Xiong · 2021 IEEE 21st International Conference on Software Quality, Reliability and Security (QRS) · 2021
Platform software reliability is critical for cloud service continuity. In this paper, citing one ultra-large-scale platform software - Intel Media Driver as an example, based on the root cause analysis to over 800 cloud reliability issues, we classify the main software reliability challenges in cloud as unexpected inputs and unexpected operation conditions. Accordingly, design perspective, we introduce one fail-safe software design against layered safeguards and validation perspective, we create the chaos driven reliability validation methodology against the calling-chain depth based grey box fuzzing. It helps us achieve 10 times reduction on the new reliability issues as well as 45% MTBF increase in two typical cloud gaming services. We expect the identified challenges can help more researchers to form the corresponding research agendas and the experiences shared can help more practitioners to apply similar enhancements.