[Edge-computing] 答复: 答复: 答复: Clarification_of_Requirements

Fu Qiao fuqiao at chinamobile.com
Wed Jun 20 02:11:59 UTC 2018


Hi, all. I guess some more explanations for the China Mobile’s network should be added here.

The definition for small and medium edge from the whitepaper actually is different from what we say about access/county/city. Our access level CO will include compute nodes of about ten more, with distance from the Base station of 10km, delay about 2ms, and bandwidth about 50GB. Our county level CO includes tens of compute nodes, with distance of 50km, delay about 2.5ms, and bandwidth of about 100GB. City level CO will have hundreds of servers, with distance of more than 100km. 

I guess the small edge defined in the Whitepaper, with up to 1U, is somehow suitable for edge equipment at the customer site, and this is not actually discussed in my presentation.

 

 

 

发件人: 赵奇慧 [mailto:zhaoqihui at chinamobile.com] 
发送时间: 2018年6月20日 10:03
收件人: Csatari, Gergely (Nokia - HU/Budapest) <gergely.csatari at nokia.com>; lebre.adrien <lebre.adrien at free.fr>
抄送: edge-computing <edge-computing at lists.openstack.org>; paul-andre raymond <paul-andre.raymond at b-yond.com>
主题: Re: [Edge-computing] 答复: 答复: Clarification_of_Requirements

 

Hi Gergely,

 

It's great that our data can help. But I think according the the previous definitions of small edge and medium edge, there should be a little modification here.

 

For Small edge, the maximum hardware specs is up to 1 unit, which is similar to our Access-level DC/Edge TIC AP. For Mediun edge, the hardware spec varies from 2RU to 20RU, which is somehow like our County-level DC. So here is my suggestion:

---------------------------------------------------------------------------

Small edge:

·        Number of instances (Edge TIC AP according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): depend on demands

·        Distance from Base Station (Access-level DC according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): around 10km 

·        E2E delay (from UE to site)(Access according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): 2 ms

·        Bandwith need (Access according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): 50 GB

 

Medium edge:

·        Number of instances (Edge TIC County according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): 3000+

·        Distance from Base Station (Access-level DC according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): around 50km

·         E2E delay (from UE to site) (Access according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): less than 2.5 ms

·        Bandwith need (Access according to  <https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#cite_note-Edge_TIC:_Future_edge_cloud_for_China_Mobile-6> [6]): 100GB

----------------------------------------------------------------------------

 

Best Regards,

Qihui Zhao

  _____  

China Mobile Research Institute

Department of Network Technology

 

发件人: Csatari, Gergely (Nokia - HU/Budapest) <mailto:gergely.csatari at nokia.com> 

时间: 2018/06/15(星期五)22:14

收件人: lebre.adrien <mailto:lebre.adrien at free.fr> ;

抄送人: paul-andre raymond <mailto:paul-andre.raymond at b-yond.com> ;edge-computing <mailto:edge-computing at lists.openstack.org> ;

主题: Re: [Edge-computing] 答复: 答复: Clarification_of_Requirements

Hi, 

I was thinking about the generic deployment specific requirements, like bandwiths and delays. 

I've added some of them from the CM presentation to https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#Small_edge and https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#Medium_edge

Br, 
Gerg0

-----Original Message-----
From: lebre.adrien at free.fr <mailto:lebre.adrien at free.fr>  [mailto:lebre.adrien at free.fr] 
Sent: Thursday, June 14, 2018 11:39 PM
To: Csatari, Gergely (Nokia - HU/Budapest) <gergely.csatari at nokia.com <mailto:gergely.csatari at nokia.com> >
Cc: Arkady Kanevsky <Arkady.Kanevsky at dell.com <mailto:Arkady.Kanevsky at dell.com> >; fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> ; paul-andre raymond <paul-andre.raymond at b-yond.com <mailto:paul-andre.raymond at b-yond.com> >; edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
Subject: Re: [Edge-computing] 答复: 答复: Clarification of Requirements

Hi all, 

Maybe we can try to prioritise the projects we want to investigate and then for each of them identify what are the missing elements/capabilities (one after the others)

There are ongoing works on Keystone and Glance. 
Nova, Neutron could be the next ones because these 4 services would allow to deploy VM at the edge (and so containers inside VMs).
Cinder will enable the use of remote attached volumes. 
Ironic can be useful later. 
etc..

My two cents,
Ad_ri3n_

----- Mail original -----
> De: "Gergely Csatari (Nokia - HU/Budapest)" 
> <gergely.csatari at nokia.com <mailto:gergely.csatari at nokia.com> >
> À: "Arkady Kanevsky" <Arkady.Kanevsky at dell.com <mailto:Arkady.Kanevsky at dell.com> >, fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> , "paul-andre raymond"
> <paul-andre.raymond at b-yond.com <mailto:paul-andre.raymond at b-yond.com> >, "lebre adrien" 
> <lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> >, edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> Envoyé: Mercredi 13 Juin 2018 17:38:35
> Objet: RE: [Edge-computing] 答复: 答复: Clarification of Requirements
> 
> Hi,
> 
> Somehow I have a feeling that these latency requirements are related 
> to all projects, this is why they should be documented in the 
> Deployment options section.
> 
> Br,
> Gerg0
> 
> -----Original Message-----
> From: Arkady.Kanevsky at dell.com <mailto:Arkady.Kanevsky at dell.com>  [mailto:Arkady.Kanevsky at dell.com]
> Sent: Wednesday, June 13, 2018 4:24 PM
> To: Csatari, Gergely (Nokia - HU/Budapest) 
> <gergely.csatari at nokia.com <mailto:gergely.csatari at nokia.com> >; fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> ; 
> paul-andre.raymond at b-yond.com <mailto:paul-andre.raymond at b-yond.com> ; lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> ; 
> edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> Subject: RE: [Edge-computing] 答复: 答复: Clarification of Requirements
> 
> Gerg0,
> I think these 3 projects are implied from the wiki requirements.
> But it will be good to state projects that may need work explicitly.
> Thanks,
> Arkady
> 
> -----Original Message-----
> From: Csatari, Gergely (Nokia - HU/Budapest) 
> [mailto:gergely.csatari at nokia.com]
> Sent: Wednesday, June 13, 2018 4:23 AM
> To: Kanevsky, Arkady; fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> ; 
> paul-andre.raymond at b-yond.com <mailto:paul-andre.raymond at b-yond.com> ; lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> ; 
> edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> Subject: RE: [Edge-computing] 答复: 答复: Clarification of Requirements
> 
> Hi,
> 
> Should we add these to the Deployment Scenarions section of the Dublin 
> wiki [1]?
> 
> [1]:
> https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#
> Deployment_Scenarios
> 
> Br,
> Gerg0
> 
> -----Original Message-----
> From: Arkady.Kanevsky at dell.com <mailto:Arkady.Kanevsky at dell.com>  [mailto:Arkady.Kanevsky at dell.com]
> Sent: Wednesday, June 6, 2018 4:49 PM
> To: fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> ; paul-andre.raymond at b-yond.com <mailto:paul-andre.raymond at b-yond.com> ; 
> lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> ; edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> Subject: Re: [Edge-computing] 答复: 答复: Clarification of Requirements
> 
> It is more than just nova to keystone.
> We also need to consider at least neutron, glance and cinder also.
> 
> -----Original Message-----
> From: Fu Qiao [mailto:fuqiao at chinamobile.com]
> Sent: Wednesday, June 6, 2018 8:21 AM
> To: 'Paul-Andre Raymond'; lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> ; 
> edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> Subject: [Edge-computing] 答复: 答复: Clarification of Requirements
> 
> Yes, 5ms is one way. But this is an assumption based on the network 
> from China Mobile. The latency will be defer if you have different 
> distance, but the calculation method is the same apparently.
> 
> -----邮件原件-----
> 发件人: Paul-Andre Raymond [mailto:paul-andre.raymond at b-yond.com]
> 发送时间: 2018年6月6日 21:19
> 收件人: Fu Qiao <fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> >; lebre.adrien at free.fr <mailto:lebre.adrien at free.fr> ; 
> edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> 主题: Re: [Edge-computing] 答复: Clarification of Requirements
> 
> Should we separate two kinds of latency requirements:
> - Federation Latency: i.e Central Keystone to Local Keystone
> - API latency: i.e. Edge Nova to local Keystone
> 
> Should we measure it one way or Round Trip? I assume the 5ms below is 
> one way.
> 
> 
> Paul-André
> --
> 
> 
> On 6/6/18, 5:13 AM, "Fu Qiao" <fuqiao at chinamobile.com <mailto:fuqiao at chinamobile.com> > wrote:
> 
> Thank you Adrien. I was just about the reply with more details.
> 
> About latency, this as I understand is actually decided mostly by
> the distance of the distributed cloud. So it actually decided by
> where exactly the location Keystone would like to deploy, and
> what is the distance expectation. Like what I explain in my
> presentation, we plan to have keystone sitting in the city level
> to control multi cloud in counties, and the latency will be
> around 5ms. But again this is a certain situation for China
> Mobile. And other operators may make the conlusion on a
> different structure. Another thing we can do is work on
> simulation and testing and see what kind of latency the current
> keystone federation scheme can tolerant. This will help the
> operators to work out there structure as well.
> 
> About bandwidth, the impression for me is we could expect more
> than 50GB of bandwidth for edge for 5G. And I think that is
> enough for most of the app.
> 
> Hope this will help.
> 
> -----邮件原件-----
> 发件人: lebre.adrien at free.fr <mailto:lebre.adrien at free.fr>  [mailto:lebre.adrien at free.fr]
> 发送时间: 2018年6月6日 15:25
> 收件人: edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> 主题: Re: [Edge-computing] Clarification of Requirements
> 
> It is rather difficult to give numbers because there are several
> use-cases.
> However, a good starting point can be to give a look to the
> presentation Qiao Fu gave during the Vancouver summit:
> https://www.openstack.org/videos/vancouver-2018/edge-tic-future-edge-cloud-for-china-mobile
> There is a lot of numbers regarding the infrastructure China
> Mobile is envisioning.
> 
> Hope this helps.
> ad_ri3n_
> PS: I cannot attend the meeting yesterday unfortunately but I'm
> wondering whether the disconnection aspects have been discussed
> (i.e. the fact that one site can be completely isolated for a
> certain period of time due to network disconnections).
> 
> ----- Mail original -----
> > De: "Jess Lampe" <jess.lampe at gmail.com <mailto:jess.lampe at gmail.com> >
> > À: edge-computing at lists.openstack.org <mailto:edge-computing at lists.openstack.org> 
> > Envoyé: Mercredi 6 Juin 2018 07:00:31
> > Objet: [Edge-computing] Clarification of Requirements
> > 
> > 
> > 
> > 
> > During the call today, members of the Glance and Keystone teams
> > requested clarity on the following areas:
> > 
> > 
> > 
> > * Latency - what are the specific latency requirements that
> > need
> > to be met?
> > * Bandwidth towards the edge - similarly, what are the
> > limitations of bandwidth at the edge that we can expect?
> > * Security - what are the specific security considerations
> > that
> > need to be?
> > 
> > 
> > Please feel free to A.) contribute additional areas that need
> > clarifying B.) clarify any of the added.
> > 
> > 
> > 
> > 
> > _______________________________________________
> > Edge-computing mailing list
> > Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing
> >
> 
> _______________________________________________
> Edge-computing mailing list
> Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
> http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing
> 
> 
> 
> 
> _______________________________________________
> Edge-computing mailing list
> Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
> http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Edge-computing mailing list
> Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
> http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing
> _______________________________________________
> Edge-computing mailing list
> Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
> http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing
>
_______________________________________________
Edge-computing mailing list
Edge-computing at lists.openstack.org <mailto:Edge-computing at lists.openstack.org> 
http://lists.openstack.org/cgi-bin/mailman/listinfo/edge-computing

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/edge-computing/attachments/20180620/c7601241/attachment-0001.html>


More information about the Edge-computing mailing list