Datacentre failover design - recommendations

[CliveG]

I have read a previous document with relevance to this, however that document is from an ISP perspective and I need this from a Customer perspective. I will attach a diagram to show the requirements:

Note: This is how it will be when completed:

3 x Datacentre in a triangle topology as described below:

DC1 - HQ DC in a VSS

DC2 - Telehouse-North - VSS

DC3 - Telehouse-East - VSS

DC1 - Port-channel to DC2 and DC3

DC2 - Port-Channel via X-Connect to DC3 - Port-Channel Upstream to BGP Collector - Port-Channel Upstream to ISP

DC3 - Port-Channel via X-connect to DC2 - Port-Channel Upstream to BGP Collector - Port-Channel Upstream to ISP

The ISPs at DC2 and DC3 are different. However, the BGP Collector to a Member LAN is the same IP on the same subnet.

There will also be a backup ADSL line to both DC2 and DC3, connected to a Console Server for remote access in case main lines fail or ports fail.

I need all routing to go via Telehouse-North and if there is an issue I need that routing to failover to Telehouse-East and then fail back to Telehouse-North once the issue has been rectified (or remain at THE - Does not really matter as long as failover works).

DC1 would also recognise this upstream failure and ammend the routing tables to route via the other port-channel to THE.

Routing protocols are:

MPLS

IS-IS

Other protocols:

Multicast for streaming traffic

The current configuration is that all traffic routes to THN. There is no upsrteam connectivy to an ISP as yet (being worked on) in THE. So, if we lose THN we lose everything. But from a routing perspective there is no built in functionality for failover.

So, when a second ISP and the Collector are connected at THE and the same routes are advertised to us, I need to ensure correct routing is in place.

Any suggestions please?

[balaji.bandi]

This required some Local understand, what applicaiton you looking DR functionality what kind of downtime you looking when the site fail over take place.

Personally i would suggest to contact local solution partner to valdidate this requirement and deploy for you.

so consultant can understand if and but's, make a working solution.

[CliveG]

Yes, understood. I am working with partners.

It's more a routing issue. I can complete this in Juniper where we say to the traffic "go this way, but if there is a failure then go the other way".

That is it really. It's so that we don't get any weird routing occurring when receiving the same routing advertisements from the upstream ISPs.