Hello, Fuqiao,

Thanks for the picture, the requirements are very clear.

I think it's OK to have one of DCs in counties working as DR/HA site. But the 
disaster recovery for OpenStack control plane is not very useful, because it's 
not easy for the control plane of OpenStack in disaster recovery site to manage 
the compute nodes in city, in this assumption, compute nodes have to be visible 
in both city OpenStack and disaster recovery OpenStack.

The more viable way of  disaster recovery(HA) is to make VNF can do the 
disaster recovery(HA) between city OpenStack and county OpenStack. Once the VNF 
in city OpenStack is in disaster status, the VNF in the county OpenStack can 
take over the traffic. It's difficult to have one to one mapping of OpenStack 
controller data in the database /VMs running on compute nodes between city 
OpenStack and county OpenStack.

"we can have Openstack deployed in the city DC, but can manage the cloud in the 
counties as well", do you mean use one OpenStack to manage the compute nodes in 
city DC and counties DCs? This mode will not save human resources and reduce 
cost. The complexity is how to manage the upgrading/patch if you have one 
OpenStack covering both city DC and counties DCs, you may have to do it all at 
the same time, and trouble shooting become more complex because the RPC 
messages spanning across city DC and counties DCs.  On the other hand, the 
DB/message bus issue will affect both city DC and counties DCs. Seems one 
OpenStack is simpler, but on contrast, it'll lead to multi-DC management and 
maintenance more complex.

Consider that each DC has almost 50 servers, I recommend to install OpenStack 
in each city DC and counties DCs, maintenance and trouble shooting(upgrading, 
patching, DB/message bus, etc) could be isolated between DCs. If you want to 
save human resources, you can have same team in city to remote manage  
OpenStack in each city DC and counties DCs, no need to have one local on-site 
team.  In fact, we almost always work remotely for our OpenStack environment in 
the lab, only go to the lab if necessary.

We can discuss it in more detail in today's weekly meeting.

Best Regards
Chaoyi Huang (joehuang)
________________________________
From: Fu Qiao [[email protected]]
Sent: 06 April 2017 10:18
To: joehuang; 'opnfv-tech-discuss'
Subject: 答复: [opnfv-tech-discuss] [multisite]weekly meeting of Apr.6

Hi, Jeo and all. Sorry for this late email about the Multisite scenario I 
mentioned in the last meeting.
Below is a very simple pic. Basically, we will have two type of DC. Core DC are 
designed to contain mainly control plane VNF, while Regional DC are designed to 
contain Data plane VNFs. For the Regional DC, we may have two type, one is DC 
in cities, with larger scale of servers(about 100), and DC in counties with 
limited number of servers( about 100).
We may require multisite support for DR for the DC in cities. Meanwhile, some 
important VNFs may require multisite HA in City DC. We are considering whether 
it is possible to have one of the DC in counties also act as the DR/HA site for 
the city DC.
In the meantime, we are also considering if openstack support cross DC 
management. We are thinking if we can have Openstack deployed in the city DC, 
but can manage the cloud in the counties as well. Such design is intend to 
reduce the human resource and cost needed in the counties.
[cid:[email protected]]

发件人: [email protected] 
[mailto:[email protected]] 代表 joehuang
发送时间: 2017年4月5日 19:24
收件人: opnfv-tech-discuss
主题: Re: [opnfv-tech-discuss] [multisite]weekly meeting of Apr.6

Hello,

For the PoC, a draft is prepared in the google doc, let's discuss and enhance 
the design

https://docs.google.com/presentation/d/16Qfm3D1yhrW31ZEVRYea8mcyKI06fQ3a6Okk4VhFGkQ/edit#slide=id.p

If you want to edit the doc, please send me your gmail account(I am not sure 
other email account can edit or not).

Best Regards
Chaoyi Huang (joehuang)
________________________________
From: 
[email protected]<mailto:[email protected]>
 [[email protected]] on behalf of joehuang 
[[email protected]]
Sent: 05 April 2017 17:11
To: opnfv-tech-discuss
Subject: [opnfv-tech-discuss] [multisite]weekly meeting of Apr.6

Hello, team,



Apr.6 2017 Agenda:

               * E release plan and OPNFV Beijing summit  discussion

* CMCC multi-site requirements discussion(2)

* Open discussion


IRC: http://webchat.freenode.net/?channels=opnfv-meeting 8:00-9:00 UTC (During 
winter time, means CET 9:00 AM).

Other topics are also welcome in the weekly meeting, please reply in this mail.

Best Regards
Chaoyi Huang (joehuang)
_______________________________________________
opnfv-tech-discuss mailing list
[email protected]
https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss

Reply via email to