Discussion of BGP as a load balancing - DR solution

[snail945]

Interested in feedback on using BGP as a Disaster Recovery and/or Load Balancing solution:

BACKGROUND

Currently, we have one production datacenter (COLO) (DC1) with 2 100MB feeds to redundant 7200s, 2 ip class C blocks (say 1.1.1.x and 2.2.2.x) one behind each

7200. this is a Cisco failover setup with all Cisco gear - mirrored 6509s., local directors, PIXs, etc). primary and secondary DNS are behind 7200 and

a firewall on a DMZ. Third dns server is at seperate corp site (CORP) on DMZ which is connected via backend T1 on another VLAN... way behind the 7200s.

Anyway, to the fun part:

OBJECTIVE:

Goal is to bring up a disaster recovery data-center (DC2) (another location/provider) where by we could route traffic to this new site should

production site go down (within an hour). It doesn't have to be utilized normally for complete loadbalancing, as it won't have all the hardware, redundancy, etc. that DC1 has. This site would have its own class C block (say 3.3.3.x) allocated from this new hosting center/ISP with backend T1 to corp and perhaps a backend T1 to product DC1 for incremental DB replication/administration, etc.

I'm a relative newbie with BGP, but based on my understanding we were thinking that BGP would help solve this problem by creating one AS comprised

of the IP blocks at both locations (1&2 from DC1 and 3 from DC2). We could inject weighted static routes into each ISPs AS respectively and if the

primary site failed traffic on the internet destined for both 1.1.1.x and 2.2.2.x would be routed over to DC2. I realize that there are many more details regarding the BGP setup but I'm trying to narrow down the functional - high level architecture to communicate internally for project approval. Is that a correct understanding? That is, can BGP function this way? I'm wondering if anyone else out there is doing this and can speak to wether (or not) BGP can help us out.

Appreciate anyone's ideas-

Byron

[svermill]

This may be an over-simplification of your requirement, but I'm not sure that I fully understand your objectives. If you are trying to find a way to send packets with a destination address of 1.1.1.x and/or 2.2.2.x to a network addressed as 3.3.3.x, I don't think you will have much success. Also, it isn't clear that your network is a candidate for BGP anyway. A single connection to a single ISP from any one single site doesn't typically call for BGP. Remember that BGP is primarily used to route between different autonomous systems. Only someone such as a ISP would route within a single AS, and only then as a transit type of network.

I have been involved in setting up "warm" sites such as your DC2 in the past. It can be quite a challenge when you have resources that must be accessed from the outside in. The inside out is quite simple of course. You may find that the *best* options available are less than ideal. For instance, if you used the same ISP, you could actually re-use your address space for critical servers, etc. and manually bring up the new "route" with your ISP as needed (probably won't meet your one hour requirement). But then you have the whole mess of replicating data between two boxes with the same IP don't you? So, depending on your underlying requirement, it is sometimes easier to simply redirect your outside "users" to the warm facility when it goes prime. Script files can help a little if you go that route.