[pve-devel] Virtio--Blk-Data-Plane deliver much higher I/O rates than both of its major competing hypervisors

[Alexandre DERUMIER]

>>Yes so dataplane is disabled while live migration but you still can't 
>>use I/O throttling so PVE would need to disable this if dataplane is on. 
>> 
>>Also "Image formats are not supported (qcow2, qed, etc).", Hot unplug is 
>>not supported. Block jobs (block-stream, drive-mirror, block-commit) are 
>>not supported. 



Yes, they are a lot of missing features. 
Maybe next year it'll be ok ;) Just need to wait a little more. 





----- Mail original -----

De: "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
À: "Alexandre DERUMIER" <aderumier at odiso.com> 
Cc: "Cesar Peschiera" <brain at click.com.py>, pve-devel at pve.proxmox.com 
Envoyé: Jeudi 6 Février 2014 09:24:28 
Objet: Re: [pve-devel] Virtio--Blk-Data-Plane deliver much higher I/O rates than both of its major competing hypervisors 

Am 06.02.2014 09:20, schrieb Alexandre DERUMIER: 
>>> A that was the culprit. Something without live migration isn't usable at 
>>> least to me. 
> 
> Just found that :) 
> 
> dataplane: enable virtio-blk x-data-plane=on live migration 
> 
> http://git.qemu.org/?p=qemu.git;a=commit;h=84db52d059f3296abf7783968645c4a96d21b099 
> 
> 
> So maybe it's possible now :) 

Yes so dataplane is disabled while live migration but you still can't 
use I/O throttling so PVE would need to disable this if dataplane is on. 

Also "Image formats are not supported (qcow2, qed, etc).", Hot unplug is 
not supported. Block jobs (block-stream, drive-mirror, block-commit) are 
not supported. 

So this would add so may exceptions to pve... i don't thing anybody 
wants that. 

Stefan 


> 
> ----- Mail original ----- 
> 
> De: "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
> À: "Alexandre DERUMIER" <aderumier at odiso.com> 
> Cc: "Cesar Peschiera" <brain at click.com.py>, pve-devel at pve.proxmox.com 
> Envoyé: Jeudi 6 Février 2014 09:15:10 
> Objet: Re: [pve-devel] Virtio--Blk-Data-Plane deliver much higher I/O rates than both of its major competing hypervisors 
> 
> Hi, 
> Am 06.02.2014 09:12, schrieb Alexandre DERUMIER: 
>>>> If i remember correctly there were some reasons why it is not available 
>>>> in PVE right now. 
>> 
>> I think they are no technical objection and not to difficult to implement. (I need to read the qemu doc) 
>> 
>> (But live-migration don't work with Virtio--Blk-Data-Plane) 
> 
> A that was the culprit. Something without live migration isn't usable at 
> least to me. 
> 
> Stefan 
> 
> 
>> ----- Mail original ----- 
>> 
>> De: "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
>> À: "Cesar Peschiera" <brain at click.com.py>, pve-devel at pve.proxmox.com 
>> Envoyé: Jeudi 6 Février 2014 08:37:49 
>> Objet: Re: [pve-devel] Virtio--Blk-Data-Plane deliver much higher I/O rates than both of its major competing hypervisors 
>> 
>> Yes that's why i'm always staring at Virtio--Blk-Data-Plane but isn't 
>> it unstable right now and has a lot of missing features? 
>> 
>> If i remember correctly there were some reasons why it is not available 
>> in PVE right now. 
>> 
>> Stefan 
>> 
>> Am 06.02.2014 08:11, schrieb Cesar Peschiera: 
>>> Hi people 
>>> 
>>> Only as comment for you (if you don't know it), I bring a link PDF of IBM 
>>> that say between many things: 
>>> 
>>> KVM with Virtio--Blk-Data-Plane deliver much higher I/O rates than both 
>>> of its major competing hypervisors 
>>> 
>>> KVM with virtio-blk technology could achieve only about 150,000 I/O 
>>> operations per second for a single guest 
>>> 
>>> KVM with Virtio--Blk-Data-Plane: (Technology Preview in RHEL 6.4) more than 
>>> 1.5 million IOPS for random I/O requests that were 4KB or less 
>>> - 49% higher than VMware vSphere 5.1 
>>> - The highest storage I/O performance ever reported in a virtualized 
>>> environment 
>>> 
>>> VMware vSphere 5.1: recently indicated that it could achieve almost 1.1 
>>> million IOPS 
>>> 
>>> In addition, KVM with virtio-blk-data-plane technology can achieve much 
>>> higher I/O rates than what Microsoft has claimed for its Hyper-V hypervisor 
>>> across multiple block sizes 
>>> 
>>> ftp://public.dhe.ibm.com/linux/pdfs/KVM_Virtualized_IO_Performance_Paper_v2.pdf 
>>> 
>>> 
>>> Best regards 
>>> Cesar 
>>> 
>>> _______________________________________________ 
>>> pve-devel mailing list 
>>> pve-devel at pve.proxmox.com 
>>> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
>> _______________________________________________ 
>> pve-devel mailing list 
>> pve-devel at pve.proxmox.com 
>> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
>> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.proxmox.com/pipermail/pve-devel/attachments/20140206/486151a6/attachment.htm>