Bug 751103

Summary: Kernel 2.6.40.8-4.fc15.x86_64 performance grossly subpar
Product: [Fedora] Fedora Reporter: Paul Lambert <eb30750>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED CANTFIX QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: urgent Docs Contact:
Priority: unspecified    
Version: 15CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2011-11-03 17:35:40 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:

Description Paul Lambert 2011-11-03 14:30:49 UTC
Description of problem:  The 8-4 kernel performs much worse that the 8-2.  None of the FE15 kernels have performed as good as those associated with FE13 and FE14.  By this I mean that these kernels could not support a virtual backend office configuration where user interaction is imperative.  These kernels are herky jerky meaning they provide a momentary burst of cpu time to the virtual machine and the put it to sleep.  There are significant delays in responding to user clicks.  And, my cpu is an AMD-V chip.  Fedora continues to flounder in a suboptimal crippled state as it never improves.  Something gets fixed in a new release but something else gets broken.  It is obvious to me that Fedora is losing its test base because computer users want to move forward not backwards.  There are so many bugs that are not being addressed while the Fedora team ignores these only to tweak the kernel so it performs worse than the previous release.  I will not be booting 8-4.  Try again.


Version-Release number of selected component (if applicable):
2.6.40.8-4.fc15.x86_64

How reproducible:
Every boot.

Steps to Reproduce:
1.Depress the computer start button to initiate a boot
2.Select kernel 2.6.40.8-4.fc15.x86_64
3.Login and start a virtual machine.
  
Actual results:


Expected results:


Additional info:

Comment 1 Dave Jones 2011-11-03 14:59:44 UTC
Is this with vmware, as in bug 730581 ?

Comment 2 Paul Lambert 2011-11-03 16:22:17 UTC
It does not matter what VM engine I am using, the point is kernel 8-2 was much improved over 7-3 and 8-4.  Kernel 8-2 was working about the same as the kernels in FE13 and FE14.  VMware works quite well with kernel 8-2 but none of the other FE15 kernels.

Comment 3 Dave Jones 2011-11-03 17:35:40 UTC
There is nothing between 8-2 and 8-4 that could explain this.
Removal of a printk, and dropping of a patch that only affected booting certain machines.

I suggest bringing it up with vmware, as there's nothing we've changed that can explain your problem.

Comment 4 Paul Lambert 2011-11-03 21:11:02 UTC
There was no graceful way for me to return to the 8-2 kernel as it is a pre-release of Verne.  Therefore, I ran preupgrade and installed the pre-release of Verne in its entirety.  Preupgrade has improved considerably and all of the needed software was installed before a reboot is necessary.  Upon reboot I was running the 8-2 kernel which is where I wanted to be.  Upon running yum update on the update-testing repo the response was no packages marked for update which leads to wonder just where the 8-4 kernel coming from.  

With the 8-2 kernel the VM engine performs better but still not as smooth as previous kernels.  For example it is much more user responsive to dragging the mouse and refreshing windows of the virtual machine but when dragging a scrollbar, it will scroll only in short bursts before freezing momentarily and then continuing.  It is my observation that when I have the VM engine in the foreground the kernel is not giving the VM engine the cpu time that it should since when doing so the Fedora user screen is in the background.  I use workspace #1 for the Fedora user interface and workspace #2 for the VMware Windows XP user interface.  The system is considerably more responsive than with the 8-4 configuration.

VMware has not released a new verson of VM Player for well over a year.  They have released some minor updates the last in the spring where it was necessary for them to delete some references to outdated kernel functions.  This was minor as these calls were already deprecated and the existing VMware release would compile and work just fine by removing the offending entries.  As I stated VMware did provide a minor update shortly after the Linux 3-39 kernel release and all was well.

Comment 5 Paul Lambert 2011-11-24 04:36:43 UTC
This new kernel, 3.1.2-1.fc16.x86_64, has improved VM performance greatly.  This is by far the best kernel of the FE16 strain.  Now when I have both a Linux desktop open and a VMware Windows open my dual cpus run at around 35%.  When I perform some activity on the VM both cpus will top out around 90%.  The VM desktop flows smoothly and the mouse/window clicks show no delay.  With the previous FE16 kernels the same activities would only drive the dual cpus up to 50%. There was some type of bottleneck limiting the cpu cycles from getting to the VM.