Automated Feature Performance Modelling of Inode Cache 
Eshwari A Madappa1, Swathi S2, Mayank Agrawal3, Suresh Anjaneyalu4
1Eshwari A Madappa*, Assistant Professor, Electronics and Communication Engineering, JSS Science and Technological University, Mysuru, Karnataka, India.
2Swathi S, M.Tech, Networking and Internet Engineering, JSS Science and Technological University, Mysuru, Karnataka, India.
2Mayank Agrawal, Member of Technical Staff at NetApp, Bangalore, Karnataka, India.
3Suresh Anjaneyalu, Member of Technical Staff at NetApp, Sunnyvale, US.
Manuscript received on June 01, 2020. | Revised Manuscript received on June 08, 2020. | Manuscript published on June 30, 2020. | PP: 1104-1108 | Volume-9 Issue-5, June 2020. | Retrieval Number: E1080069520/2020©BEIESP | DOI: 10.35940/ijeat.E1080.069520
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Inode is one of the subsystems of WAFL(Write Anywhere File Layout) file system. Inode cache is a dynamic subsystem that is percentage factor of available memory. Based on different workflows and the datasets inode cache grows and shrinks. Based on the study of customer related issues it is found that deploying such a workload and datasets at the scale, that customers typically deploy and exercise inode cache for the whole duration of test is very challenging, considering quality assurance test typically focuses on multiple subsystems. Inode cache behavior differs with steady state versus performance disruptive workflows such as volume offline, volume online, volume migration and backup/vault use cases. Based on the behavior observed on the internal test systems it is found inode cache disruptive workflows are exercised only during certain stages but not repeatedly for the duration of test and also it is hard to find out which volume is experiencing performance issues due to inode cache invalidation/shrink/rewarning. In this paper, trying to exercise the performance behavior of inode subsystem like the way customer does and try to monitor and model the subsystem using automation. Here considering the different key attributes and typical operations that effect the inode cache behavior and some of the interested counters statistics that need to be monitored for analyzing the performance behavior of inode cache. Exercising inode cache operations requires constant focus on how the inode cache is performing. Repeat and Rerun some of the targeted workflows for inode cache population invalidation/ shrink operations at constant intervals to model the behavior of the inode subsytem.
Keywords: Inode Cache Grow, Inode Cache Invalidation, Inode Cache Shrink, Snap Mirror, Volume Migration (Move).