Closed siavash79 closed 3 years ago
Hi @siavash79 ,
Are you able to reproduce the issue with the C mirror / memfs samples ? If yes, could you reopen the issue in the dokany repository since it is related to the kernel regarding your observations. Otherwise, are you seeing a flood of requests ? Do they look the same ? (spam gets info to a file, small reads...)
Thanks for you quick response
The reason I'm using Dokan .Net, is that unfortunately I'm not that experienced to use C. (I first considered microsoft projfs, but native C doesn't look like the field I can code in)
However, during the time between two posts, I did build from Github repo instead of nuget package. The result is that the cpu/kernel time has reduced considerably (although the versions look to be the same?)
However, the resource usage still seems high to me.
Regarding the spam reads: I can see from the tracelog that millions of reads are being reported, But I'm not sure if this is the intended behavior of exeplorer.exe (maybe to refresh all the time?): without using explorer, the tracelog looks much cleaner to me.
@siavash79 The idea of testing with the C samples was just to see if this is your implementation or the c# the issue. Sometimes when a value is not the one expected by Windows, it start to create more traffic and can result in this CPU usage. Like I already saw a flood of 1 byte read.
What is high resource usage for you ?
At the moment, not running Dokan.net, I have 7% cpu+kernel usage on a dual core processor.
While running Dokan.net, the cpu usage jumps up to 20% on idle time, and 80 to 100% on file operations.
@siavash79 This is not what I am seeing on my side. Could you use procmon and see if you do not have an application that is accessing the drive way more that it should ?
No activity, closing this.
Hi
I'm trying to implement a custom filesystem using dokanNet 1.4.0 (being much more flexible and easy to implement comparing to other solutions)
The issue I'm facing is that it puts of very high workload on CPU, which seems to be mostly kernel time.
Unfortunately, no matter what I do, it doesn't look willing to reduce the cpu usage, which also has made the filesystem seemingly slow.
Is there a way to improve it?