Abstract
The large volume of logs generated by devices are collected in a ring buffer to limit their storage footprint. Such a mechanism results in the potential loss of valuable diagnostic data due to buffer overwrites. This disclosure describes the use of on-device artificial intelligence (AI) models to automatically determine relevant entries in a ring-buffer crash log, remove irrelevant entries, and identify patterns within the relevant entries that correlate to crashes. The models generate and store summaries that are usable for crash analysis and debugging. The described integration of AI models can improve the precision of crash analysis, help identify recurring issues, detect potential fleet-wide issues (across a fleet of devices), and enable proactive resolution, thereby enhancing the overall user experience. The ability to query summarized error information can provide valuable insights into the performance, health, and stability of devices, which can be used to optimize software updates, identify common pain points, and prioritize development efforts.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Grizan, Sergey and Gurun, Selim, "On-Device Log Summarization Using Artificial Intelligence to Improve Crash Analysis", Technical Disclosure Commons, (April 25, 2024)
https://www.tdcommons.org/dpubs_series/6909