6

We're planning to implement a stretched datacenter between our two sites. The connectivity between them is a single 100mbit QinQ connection. We have about 500 users at site A and 200 users at site B. There is a very low amount of traffic being generated on the link.

Disclaimer: I won't go through the replication, as it's a different story. We have SAN replication ready in a few months. L3 routing etc is also not an issue, as we know how to do this.

The network plan (right now) is the following:

  1. VLAN 300: Site A+B - 172.16.0.0/24 - management (switches, routers, vsphere etc)
  2. VLAN 310: Site A+B - 10.50.0.0/24 - shared servers for clients at either site
  3. VLAN 311: Site A(+B) - 10.51.0.0/24 - servers for clients at site A
  4. VLAN 312: Site B(+A) - 10.52.0.0/24 - servers for clients at site B
  5. VLAN 320: Site A - 192.168.0.0/24 - clients in the site A building
  6. VLAN 321: Site B - 192.168.10.0/24 - clients in the site B building

The general idea here is that services with HA/site awareness (AD, DFS, Exchange, Lync and so on) lives in VLAN 311 for site A and VLAN 312 for site B. This is simply to ensure that all clients are using the services closest to them, to prevent a bad user experience if the L2 link between the sites should go down. We will be using a single AD domain.

Servers that do not have any kind of HA built in should be put in VLAN 310, wich is active at both sites simultaneously. This way we can do a SAN replica failover of a specific LUN and bring up the VM's at the opposite site in case of disaster.

Questions

  • Should I stop over-engineering stuff and rather do a completely flat design, with 3 VLAN's (one mgmt, one server and one client)? I'll lose site awareness, but it'll simplify the setup and management a great deal
  • Will vSphere break my heart if I try to implement two different clusters on the same logical network?
  • Will anything else break my heart? What are major gotchas/cons/prons?
Improve this question
1
  • No worries on vSphere running two clusters on the same networks. I've had several clusters running on the same set of networks (sharing VMotion, HA, and management networks), they're well behaved and keep out of each other's hair.
    – Shane Madden
    Oct 26, 2011 at 19:51

1 Answer 1

Reset to default
2

I think having site aware setup it better:

1) You don't want to vMotion or storage vMotion across a 100 Mbit link.

2) It would be beneficial for users to use the DC located in their building

3) As time progresses, more & more data will travel over your 100 Mbit link.

4) Plan on what will happen should the link be down for an extended period of time.

That said, you can implement 2 different clusters on the same logical network.

Improve this answer
1
  • 1
    Thanks. Preventing vmotion across the 100mbit is exactly why I want two different clusters (plus that the SAN's at each site will be isolated, so vmotion wouldnt function anyway).
    – pauska
    Oct 26, 2011 at 13:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .