Too much VCPUS makes domU high CPU utiliazation

Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 718 bytes --]


HI: 
 
     I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
 
    When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS, it starts quickly and eveything is fine.
 
     But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and domU-D 8 VCPUS,  later I start DomUA again,
(so we have total 34 VCPUS, plus dom0), it will takes very long time for domU-A to start ,and during the start, its CPU
utilizaiton is around 800% from xm top, after its start, it response very slow in VNC, ans CPU utiliazation keeps high.
And right after I destoy other 3 domUs, domUA CPU drops to normal.
 
   It might relate to CPU schedule, btw, my xen is 4.0.1.
 
  Any comments? 
 
      		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 1116 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: Too much VCPUS makes domU high CPU utiliazation

[Konrad Rzeszutek Wilk]

On Thu, May 19, 2011 at 12:19:11AM +0800, MaoXiaoyun wrote:
> 
> HI: 
>  
>      I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
>  
>     When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS, it starts quickly and eveything is fine.
>  
>      But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and domU-D 8 VCPUS,  later I start DomUA again,
> (so we have total 34 VCPUS, plus dom0), it will takes very long time for domU-A to start ,and during the start, its CPU
> utilizaiton is around 800% from xm top, after its start, it response very slow in VNC, ans CPU utiliazation keeps high.
> And right after I destoy other 3 domUs, domUA CPU drops to normal.

You might be missing some patches for the Dom0 irq round-robin code. Is your Dom0 2.6.32 pvops?
Look for these patches in your git log:
  xen: events: Remove redundant clear of l2i at end of round-robin loop
      xen: events: Make round-robin scan fairer by snapshotting each l2 word once only
      xen: events: Clean up round-robin evtchn scan.
      xen: events: Make last processed event channel a per-cpu variable.
      xen: events: Process event channels notifications in round-robin order.

>  
>    It might relate to CPU schedule, btw, my xen is 4.0.1.
>  
>   Any comments? 
>  
>       		 	   		  

> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 1966 bytes --]


My kernel version 
http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=ae333e97552c81ab10395ad1ffc6d6daaadb144a
 
I think those patch are inclued.
 
BTW, my irqbalance is disabled in dom0.
 
thanks

 
> Date: Wed, 18 May 2011 12:39:50 -0400
> From: konrad.wilk@oracle.com
> To: tinnycloud@hotmail.com
> CC: xen-devel@lists.xensource.com; george.dunlap@eu.citrix.com
> Subject: Re: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
> 
> On Thu, May 19, 2011 at 12:19:11AM +0800, MaoXiaoyun wrote:
> > 
> > HI: 
> > 
> > I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
> > 
> > When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS, it starts quickly and eveything is fine.
> > 
> > But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and domU-D 8 VCPUS, later I start DomUA again,
> > (so we have total 34 VCPUS, plus dom0), it will takes very long time for domU-A to start ,and during the start, its CPU
> > utilizaiton is around 800% from xm top, after its start, it response very slow in VNC, ans CPU utiliazation keeps high.
> > And right after I destoy other 3 domUs, domUA CPU drops to normal.
> 
> You might be missing some patches for the Dom0 irq round-robin code. Is your Dom0 2.6.32 pvops?
> Look for these patches in your git log:
> xen: events: Remove redundant clear of l2i at end of round-robin loop
> xen: events: Make round-robin scan fairer by snapshotting each l2 word once only
> xen: events: Clean up round-robin evtchn scan.
> xen: events: Make last processed event channel a per-cpu variable.
> xen: events: Process event channels notifications in round-robin order.
> 
> > 
> > It might relate to CPU schedule, btw, my xen is 4.0.1.
> > 
> > Any comments? 
> > 
> > 
> 
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel
> 
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 4128 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[Tian, Kevin]

>From: MaoXiaoyun
>Sent: Thursday, May 19, 2011 12:19 AM
>HI: 
> 
>     I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
> 
>    When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS, it starts quickly and eveything is fine.

does same thing happen if you launch B/C/D after A?

Thanks
Kevin

> 
>     But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and domU-D 8 VCPUS,  later I start DomUA again,
(so we have total 34 VCPUS, plus dom0), it will takes very long time for domU-A to start ,and during the start, its CPU
utilizaiton is around 800% from xm top, after its start, it response very slow in VNC, ans CPU utiliazation keeps high.
And right after I destoy other 3 domUs, domUA CPU drops to normal.
> 
>   It might relate to CPU schedule, btw, my xen is 4.0.1.
> 
>  Any comments? 
 
     

Re: Too much VCPUS makes domU high CPU utiliazation

[George Dunlap]

This happens only during boot, is that right?  It sounds like maybe
Linux is trying to use some kind of barrier synchronization and failing
because it's having a hard time getting all 16 vcpus to run at once.  If
that's the case, I think it's pretty much a Linux kernel issue; not sure
what the hypervisor can do about it.

 -George

On Wed, 2011-05-18 at 17:19 +0100, MaoXiaoyun wrote:
> HI: 
>  
>      I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
>  
>     When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS,
> it starts quickly and eveything is fine.
>  
>      But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and
> domU-D 8 VCPUS,  later I start DomUA again,
> (so we have total 34 VCPUS, plus dom0), it will takes very long time
> for domU-A to start ,and during the start, its CPU
> utilizaiton is around 800% from xm top, after its start, it response
> very slow in VNC, ans CPU utiliazation keeps high.
> And right after I destoy other 3 domUs, domUA CPU drops to normal.
>  
>    It might relate to CPU schedule, btw, my xen is 4.0.1.
>  
>   Any comments? 
>  
>      

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 6917 bytes --]



 

> From: kevin.tian@intel.com
> To: tinnycloud@hotmail.com; xen-devel@lists.xensource.com
> CC: george.dunlap@eu.citrix.com
> Date: Thu, 19 May 2011 11:39:14 +0800
> Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
> 
> >From: MaoXiaoyun
> >Sent: Thursday, May 19, 2011 12:19 AM
> >HI: 
> > 
> >     I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
> > 
> >    When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS, it starts quickly and eveything is fine.
> 
> does same thing happen if you launch B/C/D after A?
> 
 
>From test aspects, not really, all domains CPU Util are low.
It looks like this only happen in the process of domain A booting, and result in quite long time to boot.
 
One thing to correct, today even I destory B/C/D, domain A still comsume 800% CPU for quite a long time till now I am writing 
this mail.
 
Another strang thing is seems all VCPUS of domainUA, are running only on even number Physicals CPUS(that is 0, 2,4...),
where explains where CPU Util is 800%.  But I don't understand whether this is designed to. 
 
Below is the detail shedule log from serial.
Thanks.
 
(XEN) Scheduler: SMP Credit Scheduler (credit)
(XEN) info:
(XEN)   ncpus              = 16
(XEN)   master             = 0
(XEN)   credit             = 4800
(XEN)   credit balance     = 2324
(XEN)   weight             = 512
(XEN)   runq_sort          = 726199
(XEN)   default-weight     = 256
(XEN)   msecs per tick     = 10ms
(XEN)   credits per msec   = 10
(XEN)   ticks per tslice   = 3
(XEN)   ticks per acct     = 3
(XEN)   migration delay    = 0us
(XEN) idlers: 00000000,00000000,00000000,0000aaaa
(XEN) active vcpus:
(XEN)     1: [0.0] pri=0 flags=0 cpu=0 credit=50 [w=256]
(XEN)     2: [14.8] pri=-1 flags=0 cpu=8 credit=-32 [w=256]
(XEN)     3: [14.5] pri=-1 flags=0 cpu=14 credit=-188 [w=256]
(XEN)     4: [14.10] pri=-1 flags=0 cpu=6 credit=-187 [w=256]
(XEN)     5: [14.14] pri=-1 flags=0 cpu=8 credit=0 [w=256]
(XEN)     6: [14.11] pri=-1 flags=0 cpu=6 credit=0 [w=256]
(XEN)     7: [14.0] pri=-1 flags=0 cpu=4 credit=0 [w=256]
(XEN)     8: [14.15] pri=-1 flags=0 cpu=0 credit=297 [w=256]
(XEN)     9: [14.9] pri=-1 flags=0 cpu=8 credit=300 [w=256]
(XEN)    10: [14.2] pri=-1 flags=0 cpu=6 credit=300 [w=256]
(XEN)    11: [14.6] pri=-1 flags=0 cpu=4 credit=134 [w=256]
(XEN)    12: [14.3] pri=-1 flags=0 cpu=14 credit=288 [w=256]
(XEN)    13: [14.12] pri=-1 flags=0 cpu=12 credit=-83 [w=256]
(XEN)    14: [14.7] pri=-1 flags=0 cpu=2 credit=65 [w=256]
(XEN)    15: [14.4] pri=-1 flags=0 cpu=10 credit=-145 [w=256]
(XEN) sched_smt_power_savings: disabled
(XEN) NOW=0x00002C6DB4F0EF4B
(XEN) CPU[00]  sort=726199, sibling=00000000,00000000,00000000,00000101, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.13] pri=-1 flags=0 cpu=0 credit=-2 [w=256]
(XEN)     1: [14.15] pri=-1 flags=0 cpu=0 credit=297 [w=256]
(XEN)     2: [14.1] pri=-1 flags=0 cpu=0 credit=-1 [w=256]
(XEN)     3: [32767.0] pri=-64 flags=0 cpu=0
(XEN) CPU[01]  sort=726199, sibling=00000000,00000000,00000000,00000202, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.1] pri=-64 flags=0 cpu=1
(XEN) CPU[02]  sort=726199, sibling=00000000,00000000,00000000,00000404, core=00000000,00000000,00000000,00005555
(XEN)   run: [0.2] pri=0 flags=0 cpu=2 credit=-89 [w=256]
(XEN)     1: [14.7] pri=-1 flags=0 cpu=2 credit=65 [w=256]
(XEN)     2: [32767.2] pri=-64 flags=0 cpu=2
(XEN) CPU[03]  sort=726199, sibling=00000000,00000000,00000000,00000808, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.3] pri=-64 flags=0 cpu=3
(XEN) CPU[04]  sort=726199, sibling=00000000,00000000,00000000,00001010, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.6] pri=-1 flags=0 cpu=4 credit=134 [w=256]
(XEN)     1: [14.0] pri=-1 flags=0 cpu=4 credit=0 [w=256]
(XEN)     2: [32767.4] pri=-64 flags=0 cpu=4
(XEN) CPU[05]  sort=726199, sibling=00000000,00000000,00000000,00002020, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.5] pri=-64 flags=0 cpu=5
(XEN) CPU[06]  sort=726199, sibling=00000000,00000000,00000000,00004040, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.10] pri=-1 flags=0 cpu=6 credit=-187 [w=256]
(XEN)     1: [14.2] pri=-1 flags=0 cpu=6 credit=300 [w=256]
(XEN)     2: [14.11] pri=-1 flags=0 cpu=6 credit=0 [w=256]
(XEN)     3: [32767.6] pri=-64 flags=0 cpu=6
(XEN) CPU[07]  sort=726199, sibling=00000000,00000000,00000000,00008080, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.7] pri=-64 flags=0 cpu=7
(XEN) CPU[08]  sort=726199, sibling=00000000,00000000,00000000,00000101, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.8] pri=-1 flags=0 cpu=8 credit=-32 [w=256]
(XEN)     1: [14.14] pri=-1 flags=0 cpu=8 credit=0 [w=256]
(XEN)     2: [14.9] pri=-1 flags=0 cpu=8 credit=300 [w=256]
(XEN)     3: [32767.8] pri=-64 flags=0 cpu=8
(XEN) CPU[09]  sort=726199, sibling=00000000,00000000,00000000,00000202, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.9] pri=-64 flags=0 cpu=9
(XEN) CPU[10]  sort=726199, sibling=00000000,00000000,00000000,00000404, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.4] pri=-1 flags=0 cpu=10 credit=-145 [w=256]
(XEN)     1: [32767.10] pri=-64 flags=0 cpu=10
(XEN) CPU[11]  sort=726199, sibling=00000000,00000000,00000000,00000808, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.11] pri=-64 flags=0 cpu=11
(XEN) CPU[12]  sort=726199, sibling=00000000,00000000,00000000,00001010, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.12] pri=-1 flags=0 cpu=12 credit=-83 [w=256]
(XEN)     1: [32767.12] pri=-64 flags=0 cpu=12
(XEN) CPU[13]  sort=726199, sibling=00000000,00000000,00000000,00002020, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.13] pri=-64 flags=0 cpu=13
(XEN) CPU[14]  sort=726199, sibling=00000000,00000000,00000000,00004040, core=00000000,00000000,00000000,00005555
(XEN)   run: [14.5] pri=-1 flags=0 cpu=14 credit=-188 [w=256]
(XEN)     1: [14.3] pri=-1 flags=0 cpu=14 credit=288 [w=256]
(XEN)     2: [32767.14] pri=-64 flags=0 cpu=14
(XEN) CPU[15]  sort=726199, sibling=00000000,00000000,00000000,00008080, core=00000000,00000000,00000000,0000aaaa
(XEN)   run: [32767.15] pri=-64 flags=0 cpu=15

> Thanks
> Kevin
> 
> > 
> >     But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and domU-D 8 VCPUS,  later I start DomUA again,
> (so we have total 34 VCPUS, plus dom0), it will takes very long time for domU-A to start ,and during the start, its CPU
> utilizaiton is around 800% from xm top, after its start, it response very slow in VNC, ans CPU utiliazation keeps high.
> And right after I destoy other 3 domUs, domUA CPU drops to normal.
> > 
> >   It might relate to CPU schedule, btw, my xen is 4.0.1.
> > 
> >  Any comments? 
>  
>      
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 9100 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 1550 bytes --]



 

> Subject: Re: Too much VCPUS makes domU high CPU utiliazation
> From: george.dunlap@citrix.com
> To: tinnycloud@hotmail.com
> CC: xen-devel@lists.xensource.com; George.Dunlap@eu.citrix.com
> Date: Thu, 19 May 2011 17:21:12 +0100
> 
> This happens only during boot, is that right? It sounds like maybe
> Linux is trying to use some kind of barrier synchronization and failing
> because it's having a hard time getting all 16 vcpus to run at once. If
> that's the case, I think it's pretty much a Linux kernel issue; not sure
> what the hypervisor can do about it.
> 
> -George
 
>From the test result, it is. 
Are there some tools existed, and I can do some further analyze.
Thanks.

> 
> On Wed, 2011-05-18 at 17:19 +0100, MaoXiaoyun wrote:
> > HI: 
> > 
> > I have a host with 16 physical CPUS. Dom0 has 4 VCPUS.
> > 
> > When only start a domU-A(windows 2003 x86 64rc 2) with 16 VCPUS,
> > it starts quickly and eveything is fine.
> > 
> > But if I first start domU-B, which has 2VCPUS, domU-C 4VCPUS, and
> > domU-D 8 VCPUS, later I start DomUA again,
> > (so we have total 34 VCPUS, plus dom0), it will takes very long time
> > for domU-A to start ,and during the start, its CPU
> > utilizaiton is around 800% from xm top, after its start, it response
> > very slow in VNC, ans CPU utiliazation keeps high.
> > And right after I destoy other 3 domUs, domUA CPU drops to normal.
> > 
> > It might relate to CPU schedule, btw, my xen is 4.0.1.
> > 
> > Any comments? 
> > 
> > 
> 
> 
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 2037 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[Tian, Kevin]

>From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
>Sent: Friday, May 20, 2011 12:24 AM
>> 
>> does same thing happen if you launch B/C/D after A?
>> 
> 
>From test aspects, not really, all domains CPU Util are low.
>It looks like this only happen in the process of domain A booting, and result in quite long time to boot.

One possible reason is the lock contention at the boot time. If the lock holder
is preempted unexpectedly and lock waiter continues to consume cycles, the
boot progress could be slow.

Which kernel version are you using? Does a different kernel version expose 
same problem?

> 
>One thing to correct, today even I destory B/C/D, domain A still comsume 800% CPU for quite a long time till now I am writing 
>this mail.

So this phenomenon is intermittent? In your earlier mail A is back to normal
after you destroyed B/C/D, and this time the slowness continues. 

>Another strang thing is seems all VCPUS of domainUA, are running only on even number Physicals CPUS(that is 0, 2,4...),
>where explains where CPU Util is 800%.  But I don't understand whether this is designed to. 

Are there any hard limitation you added from the configuration file? Or do you
observe this weird affinity all the time or occasionally? It looks strange to me
that scheduler will consistently keep such affinity if you don't assign it explicitly.

you may want to run xenoprof to sampling domain A to see its hot points, or
use xentrace to trace VM-exit events to deduce from another angle.

Thanks
Kevin

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 5106 bytes --]



 

> From: kevin.tian@intel.com
> To: tinnycloud@hotmail.com; xen-devel@lists.xensource.com
> CC: george.dunlap@eu.citrix.com
> Date: Fri, 20 May 2011 07:29:55 +0800
> Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
> 
> >From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
> >Sent: Friday, May 20, 2011 12:24 AM
> >> 
> >> does same thing happen if you launch B/C/D after A?
> >> 
> > 
> >From test aspects, not really, all domains CPU Util are low.
> >It looks like this only happen in the process of domain A booting, and result in quite long time to boot.
> 
> One possible reason is the lock contention at the boot time. If the lock holder
> is preempted unexpectedly and lock waiter continues to consume cycles, the
> boot progress could be slow.
> 
> Which kernel version are you using? Does a different kernel version expose 
> same problem?
> 
 
At the very early, it is reported that VM with large number VCPUS result in high CPU utilization and 
very slow response, and the kernel version is 2.6.31.13.
 
I was trying to do some investigation now.  But the issue I report in this mail is found in 2.6.32.36.
 

> > 
> >One thing to correct, today even I destory B/C/D, domain A still comsume 800% CPU for quite a long time till now I am writing 
> >this mail.
> 
> So this phenomenon is intermittent? In your earlier mail A is back to normal
> after you destroyed B/C/D, and this time the slowness continues. 
> 
> >Another strang thing is seems all VCPUS of domainUA, are running only on even number Physicals CPUS(that is 0, 2,4...),
> >where explains where CPU Util is 800%.  But I don't understand whether this is designed to. 
> 
> Are there any hard limitation you added from the configuration file? Or do you
> observe this weird affinity all the time or occasionally? It looks strange to me
> that scheduler will consistently keep such affinity if you don't assign it explicitly.
> 
In fact I have twn machine for test, M1 and M2, for now, the issue only show M1.
I think I have no special configuration.After I compare the "xm info" of M1 and M2, I found
that M1 has two nr_nodes while M2 only have one.
This may explain where either all DomUA VCPUS all on even number of Physical CPUs.
 And also, this might cause guest slow booting.
What do u think of this?
 
thanks. 
 
 
=================M1 output=====================
release                : 2.6.32.36xen
version                : #1 SMP Fri May 20 15:27:10 CST 2011
machine                : x86_64
nr_cpus                : 16
nr_nodes               : 2
cores_per_socket       : 4
threads_per_core       : 2
cpu_mhz                : 2400
hw_caps                : bfebfbff:28100800:00000000:00001b40:009ce3bd:00000000:00000001:00000000
virt_caps              : hvm
total_memory           : 24567
free_memory            : 11898
node_to_cpu            : node0:0,2,4,6,8,10,12,14
                                      node1:1,3,5,7,9,11,13,15
node_to_memory      : node0:3997
                                      node1:7901
node_to_dma32_mem      : node0:2995
                                              node1:0
max_node_id         : 1
xen_major              : 4
xen_minor              : 0
xen_extra              : .1
xen_caps               : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32 hvm-3.0-x86_32p hvm-3.0-x86_64 
xen_scheduler          : credit
xen_pagesize           : 4096
platform_params        : virt_start=0xffff800000000000
xen_changeset          : unavailable
xen_commandline        : msi=1 iommu=off x2apic=off console=com1,vga com1=115200,8n1 noreboot dom0_mem=5630M dom0_max_vcpus=4 dom0_vcpus_pin cpuidle=0 cpufreq=none no-xsave
 
=================M2 output=====================
release                : 2.6.32.36xen
version                : #1 SMP Fri May 20 15:27:10 CST 2011
machine                : x86_64
nr_cpus                : 16
nr_nodes               : 1
cores_per_socket       : 4
threads_per_core       : 2
cpu_mhz                : 2400
hw_caps                : bfebfbff:28100800:00000000:00001b40:009ce3bd:00000000:00000001:00000000
virt_caps              : hvm
total_memory           : 24567
free_memory            : 11898
node_to_cpu            : node0:0-15
node_to_memory         : node0:11898
node_to_dma32_mem      : node0:2995
max_node_id            : 0
xen_major              : 4
xen_minor              : 0
xen_extra              : .1
xen_caps               : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32 hvm-3.0-x86_32p hvm-3.0-x86_64 
xen_scheduler          : credit
xen_pagesize           : 4096
platform_params        : virt_start=0xffff800000000000
xen_changeset          : unavailable
xen_commandline        : msi=1 iommu=off x2apic=off console=com1,vga com1=115200,8n1 noreboot dom0_mem=max:5630M dom0_max_vcpus=4 dom0_vcpus_pin cpuidle=0 cpufreq=none no-xsave

> you may want to run xenoprof to sampling domain A to see its hot points, or
> use xentrace to trace VM-exit events to deduce from another angle.
> 
> Thanks
> Kevin
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 9220 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 3090 bytes --]


Although I still not figure out why VCPU fall either on even or odd PCPUS only , If I explictly set "VCPU=[4~15]" in HVM configuration, VM will use all PCPUS from 4 to 15.
Also I may find the reason why guest boot so slow.
 
I think the reason is  the Number of Guest VCPU   >   the  Number of physical CPUs that the Guest can run on
In my test,  my physical has 16 PCPUS and dom0 takes 4, so for every Guest, only 12 Physical CPUs are available.
 
So, if Guest has 16 VCPUS, and only 12 Physical are available, when heavy load, there will be two or more VCPUS are queued
on one Physical CPU, and if there exists VCPU is waiting for other other VCPUS respone(such as IPI memssage), the waiting time 
would be much longer. 
 
Especially, during Guest running time, if a process inside Guest takes 16 threads to run, then it is possible each VCPU owns one 
thread, under physical,  those VCPUs still queue on PCPUS,  if there is some busy waiting code process, such as (spinlock),  
it will make Guest high CPU utilization.  If the the busy waiting code is not so frequently, we might see CPU utilization jump to 
very high and drop to low now and then. 
 
Could it be possible?
 
Many thanks.

 
> From: kevin.tian@intel.com
> To: tinnycloud@hotmail.com; xen-devel@lists.xensource.com
> CC: george.dunlap@eu.citrix.com
> Date: Fri, 20 May 2011 07:29:55 +0800
> Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
> 
> >From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
> >Sent: Friday, May 20, 2011 12:24 AM
> >> 
> >> does same thing happen if you launch B/C/D after A?
> >> 
> > 
> >From test aspects, not really, all domains CPU Util are low.
> >It looks like this only happen in the process of domain A booting, and result in quite long time to boot.
> 
> One possible reason is the lock contention at the boot time. If the lock holder
> is preempted unexpectedly and lock waiter continues to consume cycles, the
> boot progress could be slow.
> 
> Which kernel version are you using? Does a different kernel version expose 
> same problem?
> 
> > 
> >One thing to correct, today even I destory B/C/D, domain A still comsume 800% CPU for quite a long time till now I am writing 
> >this mail.
> 
> So this phenomenon is intermittent? In your earlier mail A is back to normal
> after you destroyed B/C/D, and this time the slowness continues. 
> 
> >Another strang thing is seems all VCPUS of domainUA, are running only on even number Physicals CPUS(that is 0, 2,4...),
> >where explains where CPU Util is 800%.  But I don't understand whether this is designed to. 
> 
> Are there any hard limitation you added from the configuration file? Or do you
> observe this weird affinity all the time or occasionally? It looks strange to me
> that scheduler will consistently keep such affinity if you don't assign it explicitly.
> 
> you may want to run xenoprof to sampling domain A to see its hot points, or
> use xentrace to trace VM-exit events to deduce from another angle.
> 
> Thanks
> Kevin
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 3779 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[Tian, Kevin]

>From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
>Sent: Friday, May 20, 2011 11:52 PM
>> 
>> >Another strang thing is seems all VCPUS of domainUA, are running only on even number Physicals CPUS(that is 0, 2,4...),
>> >where explains where CPU Util is 800%.  But I don't understand whether this is designed to. 
>> 
>> Are there any hard limitation you added from the configuration file? Or do you
>> observe this weird affinity all the time or occasionally? It looks strange to me
>> that scheduler will consistently keep such affinity if you don't assign it explicitly.
>> 
>In fact I have twn machine for test, M1 and M2, for now, the issue only show M1.
>I think I have no special configuration.After I compare the "xm info" of M1 and M2, I found
>that M1 has two nr_nodes while M2 only have one.
>This may explain where either all DomUA VCPUS all on even number of Physical CPUs.
> And also, this might cause guest slow booting.
>What do u think of this?
> 

do you enable NUMA on M1, which may be the reason for the affinity? Could you post the
cpu topology of M1 which may prove this point?

Thanks
Kevin

RE: Too much VCPUS makes domU high CPU utiliazation

[Tian, Kevin]

>From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
>Sent: Saturday, May 21, 2011 11:44 AM
>
>Although I still not figure out why VCPU fall either on even or odd PCPUS only , If I explictly set "VCPU=[4~15]" in HVM configuration, VM will use 
>all PCPUS from 4 to 15.

This may implicate that NUMA is enabled on your M1 and thus Xen scheduler tries to use
local memory to avoid remote access latency and that's why your domain A is affined to
a fix set of cpus

>Also I may find the reason why guest boot so slow.
> 
>I think the reason is  the Number of Guest VCPU   >   the  Number of physical CPUs that the Guest can run on
>In my test,  my physical has 16 PCPUS and dom0 takes 4, so for every Guest, only 12 Physical CPUs are available.

The scheduler in the hypervisor is designed to multiplex multiple vcpus on a single cpu,
and thus even when dom0 has 4 vcpus it doesn't mean that only the rest 12 pcpus are
available for use.
 
>So, if Guest has 16 VCPUS, and only 12 Physical are available, when heavy load, there will be two or more VCPUS are queued
>on one Physical CPU, and if there exists VCPU is waiting for other other VCPUS respone(such as IPI memssage), the waiting time 
>would be much longer. 
> 
>Especially, during Guest running time, if a process inside Guest takes 16 threads to run, then it is possible each VCPU owns one 
>thread, under physical,  those VCPUs still queue on PCPUS,  if there is some busy waiting code process, such as (spinlock),  
>it will make Guest high CPU utilization.  If the the busy waiting code is not so frequently, we might see CPU utilization jump to 
>very high and drop to low now and then. 
> 
>Could it be possible?

It's possible. As I replied in earlier thread, lock contention at boot time may slow down
the process slightly or heavily. Remember that the purpose of virtualization is to 
consolidate multiple VMs on a single platform to maximum resource utilization. Some
use cases can have N:1 (where N can be 8) consolidation ratio, and others may have
smaller ratio. There're many reasons for a given environment to scale up, and you need
capture enough trace information for the bottleneck. Some bottlenecks may be hard 
to tackle which will finally form into your business best practice, while some may be
simply improved by proper configuration change. So it's really too early to say whether
your setup is not feasible or not. You need dive into it with more details. :-)

Thanks
Kevin

RE: Too much VCPUS makes domU high CPU utiliazation

[MaoXiaoyun]

[-- Attachment #1.1: Type: text/plain, Size: 4045 bytes --]


 

> From: kevin.tian@intel.com
> To: tinnycloud@hotmail.com; xen-devel@lists.xensource.com
> CC: george.dunlap@eu.citrix.com
> Date: Sun, 22 May 2011 10:13:59 +0800
> Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
> 
> >From: MaoXiaoyun [mailto:tinnycloud@hotmail.com] 
> >Sent: Saturday, May 21, 2011 11:44 AM
> >
> >Although I still not figure out why VCPU fall either on even or odd PCPUS only , If I explictly set "VCPU=[4~15]" in HVM configuration, VM will use 
> >all PCPUS from 4 to 15.
> 
> This may implicate that NUMA is enabled on your M1 and thus Xen scheduler tries to use
> local memory to avoid remote access latency and that's why your domain A is affined to
> a fix set of cpus
> 
 
That's it. I've saw NUMA enabled for M1 in BIOS config while M2 is disabled.

> >Also I may find the reason why guest boot so slow.
> > 
> >I think the reason is  the Number of Guest VCPU   >   the  Number of physical CPUs that the Guest can run on
> >In my test,  my physical has 16 PCPUS and dom0 takes 4, so for every Guest, only 12 Physical CPUs are available.
> 
> The scheduler in the hypervisor is designed to multiplex multiple vcpus on a single cpu,
> and thus even when dom0 has 4 vcpus it doesn't mean that only the rest 12 pcpus are
> available for use.
>  
Oh, I should have explained that currently we pin dom0 VCPU in first PCPUS(add dom0_max_vcpus=4 dom0_vcpus_pin in grub)
And for Guest, we set "CPU=[4-15]" in HVM, so actually the PCPU for both dom0 and Guest VM is limited. We set this for the purpose
of ensure dom0 get better performance. 

> >So, if Guest has 16 VCPUS, and only 12 Physical are available, when heavy load, there will be two or more VCPUS are queued
> >on one Physical CPU, and if there exists VCPU is waiting for other other VCPUS respone(such as IPI memssage), the waiting time 
> >would be much longer. 
> > 
> >Especially, during Guest running time, if a process inside Guest takes 16 threads to run, then it is possible each VCPU owns one 
> >thread, under physical,  those VCPUs still queue on PCPUS,  if there is some busy waiting code process, such as (spinlock),  
> >it will make Guest high CPU utilization.  If the the busy waiting code is not so frequently, we might see CPU utilization jump to 
> >very high and drop to low now and then. 
> > 
> >Could it be possible?
> 
> It's possible. As I replied in earlier thread, lock contention at boot time may slow down
> the process slightly or heavily. Remember that the purpose of virtualization is to 
> consolidate multiple VMs on a single platform to maximum resource utilization. Some
> use cases can have N:1 (where N can be 8) consolidation ratio, and others may have
> smaller ratio. There're many reasons for a given environment to scale up, and you need
> capture enough trace information for the bottleneck. Some bottlenecks may be hard 
> to tackle which will finally form into your business best practice, while some may be
> simply improved by proper configuration change. So it's really too early to say whether
> your setup is not feasible or not. You need dive into it with more details. :-)
> 
I agree.  I did some extra tests. 
In "xm top", there is a colum "VBD_RD", means the number of Read IO, i count the time from VM start to first IO Read.
1) Guest VCPU = 16, set cpu="4-15" in HVM,  first IO shows up 85 seconds after VM start.
2) Guest VCPU = 12, set cpu="4-15" in HVM,  first IO shows up 20 seconds after VM start.
3)  Guest VCPU = 8, set cpu="4-7"   in HVM,  first IO shows up 90 seconds after VM start.
4)  Guest VCPU = 8, set cpu="4-15" in HVM,  first IO shows up 20 seconds after VM start.
5)  Guest VCPU = 16, set cpu="0-15" in HVM,  first IO shows up 23 seconds after VM start.
 
Previous I mentioned that we give first 4 Physical CPU *only* to dom0, it looks like, for larger VCPU guest, say 16, I 
shall not limit its PCPU to 12, but give all available PCPU to them.(just like test 5)

> Thanks
> Kevin
 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 4932 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: Too much VCPUS makes domU high CPU utiliazation

[Tian, Kevin]

[-- Attachment #1.1: Type: text/plain, Size: 4442 bytes --]

that's a nice catch. :)

Thanks
Kevin

From: MaoXiaoyun [mailto:tinnycloud@hotmail.com]
Sent: Sunday, May 22, 2011 3:45 PM
To: Tian, Kevin; xen devel
Cc: george.dunlap@eu.citrix.com
Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation


> From: kevin.tian@intel.com<mailto:kevin.tian@intel.com>
> To: tinnycloud@hotmail.com<mailto:tinnycloud@hotmail.com>; xen-devel@lists.xensource.com<mailto:xen-devel@lists.xensource.com>
> CC: george.dunlap@eu.citrix.com<mailto:george.dunlap@eu.citrix.com>
> Date: Sun, 22 May 2011 10:13:59 +0800
> Subject: RE: [Xen-devel] Too much VCPUS makes domU high CPU utiliazation
>
> >From: MaoXiaoyun [mailto:tinnycloud@hotmail.com]<mailto:[mailto:tinnycloud@hotmail.com]>
> >Sent: Saturday, May 21, 2011 11:44 AM
> >
> >Although I still not figure out why VCPU fall either on even or odd PCPUS only , If I explictly set "VCPU=[4~15]" in HVM configuration, VM will use
> >all PCPUS from 4 to 15.
>
> This may implicate that NUMA is enabled on your M1 and thus Xen scheduler tries to use
> local memory to avoid remote access latency and that's why your domain A is affined to
> a fix set of cpus
>

That's it. I've saw NUMA enabled for M1 in BIOS config while M2 is disabled.

> >Also I may find the reason why guest boot so slow.
> >
> >I think the reason is  the Number of Guest VCPU   >   the  Number of physical CPUs that the Guest can run on
> >In my test,  my physical has 16 PCPUS and dom0 takes 4, so for every Guest, only 12 Physical CPUs are available.
>
> The scheduler in the hypervisor is designed to multiplex multiple vcpus on a single cpu,
> and thus even when dom0 has 4 vcpus it doesn't mean that only the rest 12 pcpus are
> available for use.
>
Oh, I should have explained that currently we pin dom0 VCPU in first PCPUS(add dom0_max_vcpus=4 dom0_vcpus_pin in grub)
And for Guest, we set "CPU=[4-15]" in HVM, so actually the PCPU for both dom0 and Guest VM is limited. We set this for the purpose
of ensure dom0 get better performance.

> >So, if Guest has 16 VCPUS, and only 12 Physical are available, when heavy load, there will be two or more VCPUS are queued
> >on one Physical CPU, and if there exists VCPU is waiting for other other VCPUS respone(such as IPI memssage), the waiting time
> >would be much longer.
> >
> >Especially, during Guest running time, if a process inside Guest takes 16 threads to run, then it is possible each VCPU owns one
> >thread, under physical,  those VCPUs still queue on PCPUS,  if there is some busy waiting code process, such as (spinlock),
> >it will make Guest high CPU utilization.  If the the busy waiting code is not so frequently, we might see CPU utilization jump to
> >very high and drop to low now and then.
> >
> >Could it be possible?
>
> It's possible. As I replied in earlier thread, lock contention at boot time may slow down
> the process slightly or heavily. Remember that the purpose of virtualization is to
> consolidate multiple VMs on a single platform to maximum resource utilization. Some
> use cases can have N:1 (where N can be 8) consolidation ratio, and others may have
> smaller ratio. There're many reasons for a given environment to scale up, and you need
> capture enough trace information for the bottleneck. Some bottlenecks may be hard
> to tackle which will finally form into your business best practice, while some may be
> simply improved by proper configuration change. So it's really too early to say whether
> your setup is not feasible or not. You need dive into it with more details. :-)
>
I agree.  I did some extra tests.
In "xm top", there is a colum "VBD_RD", means the number of Read IO, i count the time from VM start to first IO Read.
1) Guest VCPU = 16, set cpu="4-15" in HVM,  first IO shows up 85 seconds after VM start.
2) Guest VCPU = 12, set cpu="4-15" in HVM,  first IO shows up 20 seconds after VM start.
3)  Guest VCPU = 8, set cpu="4-7"   in HVM,  first IO shows up 90 seconds after VM start.
4)  Guest VCPU = 8, set cpu="4-15" in HVM,  first IO shows up 20 seconds after VM start.
5)  Guest VCPU = 16, set cpu="0-15" in HVM,  first IO shows up 23 seconds after VM start.

Previous I mentioned that we give first 4 Physical CPU *only* to dom0, it looks like, for larger VCPU guest, say 16, I
shall not limit its PCPU to 12, but give all available PCPU to them.(just like test 5)

> Thanks
> Kevin

[-- Attachment #1.2: Type: text/html, Size: 8953 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel