Hello Salman,

I make the assumption there is a router on the remote site where the two DSL links are terminated.

using two different Vlans/IP subnets provide additional benefits (true QoS treatment up to L2 switch port becomes possible).

However, it is not strictly required.

The right tool to perform routing decisions on criteria different from simple destination address is called PBR= policy based routing.

Using a route-map traffic received on interface is examined and it can be diverted to one link or to the other.

So you need to define an ACL that define what is VoIP traffic including VOIP signalling and RTP bearer streams.

Example

route-map pbr_voip_data permit 10

match ip address voice_traffic

set interface fas0/0

route-map pbr_voip_data permit 20

set interface fas0/1

you still have two static routes for each IP subnet that is beyond the two WAN links

ip route 10.10.10.0 255.255.255.0 fas0/0

ip route 10.10.10.0 255.255.255.0 fas0/1

ip route 10.10.20.0 255.255.255.0 fas0/0

ip route 10.10.20.0 255.255.255.0 fas0/1

to build a smart solution that can detect failures you need to refer to reliable static routing

see

http://www.cisco.com/en/US/docs/ios/12_3/12_3x/12_3xe/feature/guide/dbackupx.html

Actually the same approach can be used to verify reachability of next-hops in the route-map used by PBR.

Edit:

if you don't have a router at remote site but only a L2 switch and terminating modems for DSL connections:

you can get separation of traffic by using two different vlans/IP subnets and using modems as bridges.

However, if one link fails the other one cannot take over.

On the other end if each link is a L2 trunk carrying both vlans you would need to use IRB on router but it doesn't support per vlan STP so you would end with one link used for both types of traffic and one for backup only.

Hope to help

Giuseppe