Best way to stand up new site

[CiscoPurpleBelt]

I  know there is a lot involved, but from a networking perspective, what is the best way to start building a new hub network/site that will replace the existing hub/site while keeping all IPs the same once we completely fail over. All remote sites and services must remain accessible. Any advice guys? 

[chrihussey]

If it is possible, establish L2 connectivity between the old and new hub sites locations. Depending on the complexity of whats involved this could greatly simplify the migration as you could have the same subnet active in both locations. This would alleviate the need to move all devices in a network over in one shot. You would probably want to build in a degree of redundancy because any loss of connectivity between the two locations during the transition period would be very messy.

Hope this helps

[CiscoPurpleBelt]

Can you give me a quick basic config of what you are talking about?

[marce1000]

 - Question is, are these requirements correct and or must it be done, that way. In the modern world apps  or services far from relate to IP's anymore. Many technologies exists to make app's failover to new underlying IP's such as domain TTL's and or MX records for mail (an example). Also  private addresses could be used 'locally' within the new HUB; firewalls may NAT to public addresses were need (etc).

M.

[chrihussey]

Yes, there are probably a million ways to skin a cat. With an undertaking such as this one could get as creative as one wants and there is probably no right or wrong answer. In the end it is results that matter. However, in my years of networking I have found simplicity works best. Just my opinion.

[CiscoPurpleBelt]

Yes I want it as simple as possible as well as too complicated always has more problems in my experience as well. Basically, we need to:
be able to access the new and old hub sites from either location to build it while current networks are still up.
All IPs must be/remain the same once finished.
What do you suggest?

[CiscoPurpleBelt]

I don't know too much about the sytems infrastructure here but I would assume MX is being used for exchange, I don't know much about the rest.
The thing is, over here we must keep the IPs the same as it would cause us to have to change IPs on devices and many remote sites. Basically, we need to:
be able to access the new and old hub sites from either location to build it while current networks are still up.
All IPs must be/remain the same once finished.
What do you suggest?

[chrihussey]

The best way and probably optimal thing would be to get dark fiber(s) between the old and new data centers. This would be contingent on distances more than anything. However, it would alleviate the need to have a provider in the middle, allow for flexible bandwidth and simplify things to a great degree. If this is possible, then the new data center would simply be an extension of the existing. So keeping all IPs the same would not be an issue.

[CiscoPurpleBelt]

When you say dark fiber you mean like using SM fiber or something? The distance between current and new site is too far - about 7 miles or so.
We could simply give the edge device at the new site a different IP so it can be accessed from current site, and have layer 2 connectivity to reach all the new servers that will be replicated at new site,but everything else on network devices over at new site can be built just the same as current site? This sound like a good way to go?

[chrihussey]

Yes, SM fiber. Many providers offer it as a service. In SM terms 7 miles is not a great distance. Switch optics can go anywhere from 10 to 80 kilometers. With a single pair of fiber between the two data centers you could easily get 10G of connectivity. There are also ways of increasing that. It would not be a bad idea to get two diversely routed pairs for redundancy. 

So to further illustrate the concept. Just as you would add a switch in your existing data center to allow more hosts/servers access to a network, adding one on the other side of the dark fiber would achieve the same results. In essence the distance between the two locations would be negligible. You would merely be extending the data center LANs to the new location.

[CiscoPurpleBelt]

If everything still needs to have the same IPs, what is the best way to prevent any conflicts?

[chrihussey · Accepted solution]

I'm sure there are some complexities, but IP conflict issues should be minimal, if any.

If a server with an IP is operating at the old data center and needs to be replaced with one at the new, then of course they cannot operate at the same time. This would need to be planned and properly executed. Failing back would be relatively easy as turning off the new and activating the old is all that would be necessary.

However, with virtualization, all this is greatly simplified and moving a server from one location to another makes this even easier.

[CiscoPurpleBelt]

Yes I still need to get with the systems guys as to find out more about the back end and infrastructure. Yes this was what I had in mind. Let's say that this is the way it had to be done, we would need to test. As long as new servers/equipment is reachable from both sites, we could test the services using different IPs then change it back to IPs just like how it is now when we plan to fail-over? Having techs locally at both sites is how things will be during the implementation phase.