Memory Degradation Analysis in Private and Public Cloud Environments

Memory degradation trends have been observed in many continuously running software systems. Applications running on cloud computing can also suffer from such memory degradation that may cause severe performance degradation or even experience a system failure. Therefore, it is essential to monitor such degradation trends and find the potential causes to provide reliable application services on cloud computing. In this paper, we consider both private and public cloud environments for deploying an image classification system and experimentally investigate the memory degradation that appeared in these environments. The degradation trends in the available memory statistics are confirmed by the Mann-Kendall test in both cloud environments. We apply causal structure discovery methods to process-level memory statistics to identify the causality of the observed memory degradations. Our analytical results identify the suspicious processes potentially leading to memory degradations in public and private cloud environments.