I have a Cisco router with two bonded T1's which are setup as a VPN to the main office. We need more bandwidth but can't get other connections (or it's too costly), so I would like to have a dsl connection installed. This DSL connection will run over a VPN to the same main office, but it won't be bonded with the T1's - so it won't act as a single connection.
Since the three circuits won't act as a single connection (basically would be two connections 2 T1's + 1 DSL) we would have to split the network in half - but I don't want to do that. Instead, would it be possible to send all HTTP/HTTPS over the DSL connection but send all mission critical data (such as voice/active directory) over the T1's?
I basically want to send specific ports over DSL and everything else over the T1's without separating half of the users traffic over the DSL and the rest over the T1's.
Depending on the make and software of your router, you should be able to use policy based routing to accomplish what you want to do. The cisco docs have info on PBR - there are other blog sites as well. This one looks close to what you want to do and may walk you through it enough that you can figure it out.
Just to state the obvious, this does nothing for any inbound services you have.
You may also want to look into load balancing device. One I've used with success in the past are the Peplink devices, but there are other ones out there. A dedicated device will provide you much more flexibility and depending on your environment, can even provide for inbound traffic failover and load balancing.
While this can be done with a router setup, the path of least resistance for you, in terms of setting policy and integrating with multiple connection types, is to use a link balancer. I make heavy use of Elfiq link balancer appliances in these situations. These days, there's no excuse for a business to only rely on a T1 or a single medium, especially when users are used to broadband speeds at home. The Elfiq and others (PepLink, Barracuda) can combine DSL, T1, Cable, 3G/4G, fiber, etc. and will allow you to set policies and persistence across connection types.
A typical client setup for me will have a T1 link as primary for inbound services, maybe another bonded T1 for additional bandwidth, and a high-speed asynchronous DSL or cable modem for all web traffic (80/443). If any line fails, the other lines are used. If only on line is active, all traffic falls back to it. If any line hits a specified threshold, say 80%, new connections are distributed among the less-busy connections.
So when you install something like this, you can view your connections as a pool of aggregated resources. It's transparent to the end-users.
You might want to give dynamic routing a closer look - EIGRP can load balance between non equal lines.
On the other hand you could use policy based routing to select your outbound interface according to acls.
tsg