Replacing T1 with Metro?

[RadiclDreamer]

I am looking at the possibility of replacing my offsite network with something a little faster.

Current setup is a DS3 at the main site with T1 lines from each of the remotes doing a point to point. Config looks like

interface Serial4/0.2 point-to-point
description T1 to SITE1
ip address 172.25.40.13 255.255.255.252
frame-relay interface-dlci 32 IETF
!
interface Serial4/0.3 point-to-point
description T1 to SITE 2
ip address 172.25.40.21 255.255.255.252
frame-relay interface-dlci 33 IETF
!
interface Serial4/0.4 point-to-point
description T1 to SITE 3
ip address 172.25.40.17 255.255.255.252
frame-relay interface-dlci 34 IETF
!
interface Serial4/0.5 point-to-point
description T1 to SITE 4
ip address 172.25.40.9 255.255.255.252
frame-relay interface-dlci 35 IETF
!
interface Serial4/0.6 point-to-point
description Bonded T1 to SITE 5
ip address 172.25.40.1 255.255.255.252
frame-relay interface-dlci 36 IETF
!
interface Serial4/0.8 point-to-point
description T1 to SITE 6
ip address 172.25.40.29 255.255.255.252
frame-relay interface-dlci 202 IETF


ETC ETC

And on the remote side

interface MFR1
description Virtual Bonded interface 2 T1
ip address 172.25.40.2 255.255.255.252
encapsulation frame-relay IETF
frame-relay interface-dlci 36
!
interface FastEthernet0/0
ip address 172.25.73.1 255.255.255.0
ip helper-address 172.25.21.1
duplex auto
speed auto
!



If i were to go with a metro connection and have 2960 switches to work with at the remote ends and a 7206VXR at the head end, what kind of setup would I be looking at?

 

[drebo]

You'd be using VLAN-tagged subinterfaces off the 7206VXR. Each site would be a VLAN.

I would strongly recommend, though, using /30 subnets between sites, which would require you to put a router at the branch sites. The reason for this is that if you don't do it, all broadcast traffic from those sites will traverse the L2 VPN, which is extra traffic that doesn't really need to.

If your pipes are fat enough, I guess it wouldn't matter.

From a refinement point of view, you'd want to configure traffic shaping outbound on each of the subinterfaces to whatever the commit rate is for the EVPL to the remote site.

 

[TheCrackLing]

Are you looking at a Point to Point from each site back to the main site? Or are you looking at a Point to Multi-Point?

If the former, then just do roughly the same thing you are doing now, except you'd have tagged VLANs for each site, and yes, definitely go router at each site just like you are doing now.

If it's the latter, then you're probably going to be setup on more than likely a MPLS L3VPN. In which case, you'd have a route from you to the carrier (which gets tagged to be only yours and can be done by any routing protocol, though typically BGP is used) at each site.

The advantage of the latter is, if you have a lot of site traffic that has to go to other sites, but not the main site, you will gain bandwidth.

Either way, it really depends on what you mean by "Metro." Honestly, I see no point in going the former router, and would always suggest the latter just because it's better, even though it's a bit more expensive from the hardware standpoint.

Also, please stop suggesting people use /30s, you should be using /31s for any and all P2P Router circuits. Save those IPs, especially if you are using public space at all.

 

[drebo]

Also, please stop suggesting people use /30s, you should be using /31s for any and all P2P Router circuits. Save those IPs, especially if you are using public space at all.

They're not using public space and if they're using a broadcast medium they should be using /30s.

Sure, for an actual non-broadcast point-to-point circuit, they can use /31s or even unnumbered links.

 

[Cooky]

I second on the MPLS L3VPN option.
It's so much easier to work w/ routed networks.
Even if you start growing into 1000 remote sites, you'd have only one interface to manage at the head-end.

Many years ago we had exactly same setup as yours (head-end hub w/ P2P frame to remote).
After migrating to MPLS our lives had become so much easier.

Basically it's got easier management, support, and better scalability.

Edit: Unless the 7206 is old/existing hardware, I'd go w/ ASR instead.
The 7200's will probably reach EoX status soon, if not already.

 

[RadiclDreamer]

You'd be using VLAN-tagged subinterfaces off the 7206VXR. Each site would be a VLAN.

I would strongly recommend, though, using /30 subnets between sites, which would require you to put a router at the branch sites. The reason for this is that if you don't do it, all broadcast traffic from those sites will traverse the L2 VPN, which is extra traffic that doesn't really need to.

If your pipes are fat enough, I guess it wouldn't matter.

From a refinement point of view, you'd want to configure traffic shaping outbound on each of the subinterfaces to whatever the commit rate is for the EVPL to the remote site.

I had thought about doing this, being that i have 1841 routers at every location I guess i could use them. Just move the current setup from the s0/0/0 interfaces to the eth and then on the 7206 side move the config from the serial4/0.x to a new eth1/0.x interface. I was hoping to drop the amount of equipment at each location but i guess this would really be the best bet.