错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

EasyLog: An Efficient Kernel Logging Service for Machine Learning

  • Xundi Yang,
  • Kefan Qiu,
  • Quanxin Zhang

摘要

Recently, logs serves as a crucial tool to monitor system’s real-time state for experiments and generate data for machine learning. However, the existing Linux logging system faces challenges such as excessive log output and a high rate of important log message loss. To tackle these issues, we propose the EasyLog solution, which effectively mitigates these problems. EasyLog draws inspiration from the design principles of log-related functions like pr_xx, dev_xx, and the Devkmsg service. EasyLog extracts and records logs with special identifier suffixes by introducing a ring buffer. In terms of interface utilization, EasyLog offers the easy_xx interface for kernel developers and the reading interface for user-space applications.