I know how to profile CPU/Memory. But now I need to profile some code which uses IO operations (PlainSocketImpl.doConnect and others) to understand which piece of code is taking most of the time.
我知道如何配置CPU /内存。但是现在我需要分析一些使用IO操作的代码（PlainSocketImpl.doConnect和其他代码），以了解大部分时间内哪些代码占用。
2 个解决方案
#1
2
JProfiler has file and sockets probes that show you I/O data around the actual objects.
JProfiler具有文件和套接字探针，可以显示实际对象周围的I / O数据。
In the CPU views there is a thread selector that has a "Net I/O" state. If you select that state, you will only see times spent in I/O.
在CPU视图中，有一个具有“Net I / O”状态的线程选择器。如果选择该状态，则只会看到在I / O中花费的时间。
Disclaimer: My company develops JProfiler.
免责声明：我公司开发JProfiler。
#2
3
On linux based systems you can try iotop. Switch the view to show threads, then combine it with jstack outputs to match actual lines of code.
在基于Linux的系统上，您可以尝试iotop。切换视图以显示线程，然后将其与jstack输出组合以匹配实际的代码行。
Not very automatic though.
虽然不是很自动。
#1
2
JProfiler has file and sockets probes that show you I/O data around the actual objects.
JProfiler具有文件和套接字探针，可以显示实际对象周围的I / O数据。
In the CPU views there is a thread selector that has a "Net I/O" state. If you select that state, you will only see times spent in I/O.
在CPU视图中，有一个具有“Net I / O”状态的线程选择器。如果选择该状态，则只会看到在I / O中花费的时间。
Disclaimer: My company develops JProfiler.
免责声明：我公司开发JProfiler。
#2
3
On linux based systems you can try iotop. Switch the view to show threads, then combine it with jstack outputs to match actual lines of code.
在基于Linux的系统上，您可以尝试iotop。切换视图以显示线程，然后将其与jstack输出组合以匹配实际的代码行。
Not very automatic though.
虽然不是很自动。