So strange, time becomes skew when VM is running

So strange, time becomes skew when VM is running

[Feng Wang]

[-- Attachment #1: Type: text/plain, Size: 1593 bytes --]

When I started up two KVM VMs upon SLES 10 OS based on a multi-core hardware
platform and ran some computing-intensive tasks (e.g. SPECjbb2005) within each VM,
I found that something confusing me very much.
The VM which is first started is running as well as it runs without other VMs, but
the system time of the subsequent started one is skew. E.g. VM1 and VM2 are both
started at 10:00am, and then I execute the same one computing-intensive task in
each VM at nearly same time, i.e. 10:15am. Then you will see that the execution of
task in VM2 is much slowly than VM1 (There are enough computing resource for VMs,
e.g. the hardware platform have 8 CPUs, and each VM is configured by supporting 1
CPUs). When the task in VM1 is finished, the system times of VM1 and Host are both
10:30am, this means the execution of task needs nearly 15 minutes. But when the
task in VM2 is finished, the system time of Host is 10:40am, the system time of
VM2 is 10:30am, so the system time of VM2 is slower than the Host's. Why?
Additionally, I found that the CPU¡¯s ¡°bogomips¡± feature listed in /proc/cpuinfo
of each VM is different. I guess this maybe results from the CPU emulation of QEMU
because I found the similar phenomenon when I used KVM VM with the ¡°-no-kvm¡±
option. I don¡¯t know if this is the reason for the problems described above, but
according to my experiment results I found that the bogomips feature has few
relations with the execution performance of VM. 
I don¡¯t know if this can demonstrate the performance conflict between two VMs, so
I need your help. Thank you very much.




[-- Attachment #2: Type: text/plain, Size: 314 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/

[-- Attachment #3: Type: text/plain, Size: 186 bytes --]

_______________________________________________
kvm-devel mailing list
kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Re: So strange, time becomes skew when VM is running

[Avi Kivity]

Feng Wang wrote:
> When I started up two KVM VMs upon SLES 10 OS based on a multi-core hardware
> platform and ran some computing-intensive tasks (e.g. SPECjbb2005) within each VM,
> I found that something confusing me very much.
> The VM which is first started is running as well as it runs without other VMs,
>   

The first VM is probably using the RTC as the time source;  
unfortunately only one process can use the RTC at a time.  If you use 
newer kvms (and newer host kernels), and configure HZ appropriately 
(dynamic tick or HZ=1000), then timing should be much better for all VMs.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/