Multi-gigabit edge devices as CPE

Hamish McGlinn hmcglinn at
Wed Apr 8 23:26:00 UTC 2015

Is it a necessity to terminate the layer 3 at the edge? You could get a
10Gbps switch and move it all back to a central location where you have
your high end routers. It would then be terminated as a VLAN and be a
router on a stick kind of topology. Could be a cheaper way to do it without
taking MPLS all the way out to the edge.

As Tim said above, I too was thinking about the Juniper ACX. The 5048/5096
model could suit your needs. They are primarily designed as layer 1(TDM)/2
backhaul devices and i'm not sure they can do a full table. They do have
full JunOS MPLS features. Could be a way to use MPLS-TE to move the layer 2
back to a core location and terminate later 3 there. Would give you some
flexibility over just doing ethernet stuff as I mentioned in the first


On Thu, Apr 9, 2015 at 10:46 AM, Daniel Rohan <drohan at> wrote:

> I work at a state REN and we are seeking a lead for a new edge device for
> on prem deployment at customer sites.
> We currently deploy two classes of routers-- a high end and a low end. Both
> the high end and the low end use some of the standard edge features:
> MPLS-TE, MBGP, flowspec, vrf, PIM, etc. We deliver full tables over these
> devices to the customers that need them.
> We recently finished a new ethernet procurement and have a large number of
> sites (~200) moving from <1Gbps in bandwidth to 1-10Gb in bandwidth. Our
> currently deployed low-end router can't handle these speeds and we can't
> afford to place our high end router at 200+ sites.
> So, we're looking for a middle tier router to deploy. Something with 2+
> SFP+ ports, software that can handle the aforementioned features, and
> something with an API that we can leverage for programmatic management.
> So far we've not found anything that checks all the boxes. Layer 3 switches
> seem like obvious choices, but lack some of the features and RIB/FIB we
> need at the edge. Other devices like the Juniper MX5/10 certainly meet the
> requirements, but are priced way beyond what we can afford.
> Any suggestions for devices we might have overlooked? Preferably in the
> less than 10K per unit price point. If such a magical device exists.
> -Dan

More information about the NANOG mailing list