발간년도 : [2022]
| 논문정보 |
|
| 논문명(한글) |
[Vol.17, No.4] Implementation of Tracing-based Net-perf for Linux Kernel Network Performance Analysis |
|
| 논문투고자 |
Ju-Hee Kang, Tae-Seong Kim, Jae-Won Shin, Yoon-Jin Ahn, Ik-Su Kim |
|
| 논문내용 |
These days network systems are getting more complex, making it much more difficult to measure performance for code modification. A typical example of this complexity is Linux because it was developed to meet all of the various networking requirements. In particular, it is truly difficult to find a tool to analyze the resulting performance changes after making changes to the Linux Kernel Networking Stack. It can be tried by measuring and comparing network speeds before and after performance improvements using a tool called iperf3 but still, the problem exists that it is not sure whether the speed improvement occurred at a specific point intended by the developer in complex configured code. To improve these problems, this paper proposes a new performance analysis tool that combines tracing. In the proposed tool, we use a technique called tracing to track the Linux kernel function call process to visualize the clear point where performance improvements have occurred. Therefore, the user can determine the exact cause of the performance improvement based on various information such as whether the function execution latency at that point has decreased or whether the function's call depth has decreased. This paper also aims to prove the usefulness of the tool by analyzing the case in which performance improvement occurred by changing the data structure of the HSR module on Linux Kernel to Hash List. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
