OK, to help us investigate it, I created a bugzilla entry for it,
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1143;
And previously, there were some discussions on this issue, I attached them
below.
Steven Hand wrote:
>> 2) HVM domain performance would downgrade, after doing save/restore.
>
> Is there any more information on this? I can't find any record of it
> as being a 'new issue' ever, just being an 'old issue' for quite a
> while now. Is there a bugzilla entry? Or can you just give more
> details via email?
=========================
For the virtual interrupt injection rate, there is almost no change across
save/restore;
The save/restore happens on the same machine.
The downgrade happens under all combinations of 32pae/32e VMX guests on
32pae/32e Xen, and it should exist for quite a long period of time.
-- Dexuan
Keir Fraser wrote:
> Does the interrupt rate change after save/restore? Perhaps we are
> creating a timer-interrupt storm immediately after restore (but this
> should subside quickly, or the performance drop should be much more
> dramatic), or we have got the restored interrupt rate wrong somehow?
> Is the save/restore on the same machine, or across machines?
>
> -- Keir
>
> On 30/12/07 11:29, "Li, Xin B" <xin.b.li@xxxxxxxxx> wrote:
>
>> I saw 32bit Vista guest becomes very slow after a save/restore if
>> using default timer_mode 0, but timer_mode 2 is OK.
>> -Xin
>>
>>> -----Original Message-----
>>> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>>> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Li, Xin
>>> B Sent: 2007年12月30日 19:23 To: Keir Fraser; Cui, Dexuan; You,
>>> Yongkang; xen-devel
>>> Subject: RE: [Xen-devel] VMX status report. Xen:#16673 & Xen0: #372
>>> -oneissuefixed
>>>
>>> will timer_mode impact guest performance after a save/restore? -Xin
>>>
>>>> -----Original Message-----
>>>> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>>>> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Keir
>>>> Fraser Sent: 2007年12月30日 0:14 To: Cui, Dexuan; You, Yongkang;
>>>> xen-devel
>>>> Subject: Re: [Xen-devel] VMX status report. Xen:#16673 & Xen0:
>>>> #372 -oneissue fixed
>>>>
>>>> Not really. I'd perhaps get Xen to dump information about all
>>>> hvm_params and acceleration options and see if anythign has been
>>>> dropped across the save/restore.
>>>>
>>>> -- Keir
>>>>
>>>> On 29/12/07 13:24, "Cui, Dexuan" <dexuan.cui@xxxxxxxxx> wrote:
>>>>
>>>>> On my host, for KernelBuild test, it may vary from 3% to 10%, but
>>>>> 30% was also observed on others' hosts.
>>>>>
>>>>> Any suggestion?
>>>>>
>>>>> Thanks
>>>>> -- Dexuan
>>>>>
>>>>> -----Original Message-----
>>>>> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
>>>>> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Keir
>>>>> Fraser Sent: Saturday, December 29, 2007 5:20 PM
>>>>> To: You, Yongkang; xen-devel
>>>>> Subject: Re: [Xen-devel] VMX status report. Xen:#16673 & Xen0:
>>>>> #372 - oneissue fixed
>>>>>
>>>>> On 29/12/07 06:40, "You, Yongkang" <yongkang.you@xxxxxxxxx> wrote:
>>>>>
>>>>>> 6) HVM domain performance would downgrade, after doing
>>>>>> save/restore.
>>>>>
>>>>> How much does it downgrade by?
>>>>>
>>>>> -- Keir
>>>>>
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
-- haicheng
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home