Comments inline.
On 9/28/06 8:02 PM, "Tom Monte" wrote:
> I am the project lead on installing MPLS to replace
some ISDN PRI/BRI and
> frame-relay as our backup lines. The initial pilot
will be ten to fifteen
> sites on one MPLS network and possible expanding to two
MPLS networks (a
DMZ
> backup and a Internal backup) with 100 sites. I have
numerous concerns
> technically about what our vendor is selling us.
>
> 1.) We are debating between Layer 2 MPLS and Layer 3
MPLS. The vendor and
> our accounting department are steering us heavily
toward Layer 3. Why
would
> they want to force us into Layer 3? We are running
EIGRP internally and
the
> benefit of Layer 2 is we can run EIGRP over the top of
it just like any
> other line. If we go to Layer 3 we add the complexity
of iBGP to our
> network. It sounds like the only benefit of Layer 3 is
CoS/Qos. Are there
> any other benefits?
You don't necessarily HAVE to run BGP with the PE routers;
it is just one of
the most popular methods of exchanging routes. You can use
static routes,
RIP, OSPF, or EIGRP. Carriers prefer to use either static
routes or BGP. I
was looking at having native EIGRP as an option for a
client, but it was
hard to find a carrier that was willing, or capable of doing
so.
The scope of the client's use of BGP in a typical MPLS VPN
network is kinda
limited; just between the CE and PE, and if your site is
dual-attached, iBGP
amongst your routers. I like this because it limits the
extension of your
IGP; each site becomes almost like an island onto itself.
In all of this, did the benefit of any-to-any connectivity
come up? It is a
major benefit for peer-to-peer traffic types, like VoIP.
>
> 2.) Will MPLS gracefully scale to 100 sites? Will
there be a big increase
> in the cpu and memory I need on my routers?
Furthermore, if I have to use
> iBGP and have 100 sites, unless I use route reflectors
or confederations,
> wont I need 100 * (100-1) connections? This sounds
ugly on a large scale.
> The vendor told me I would only need one connection to
the hub. This
> doesn't make sense to me.
MPLS will easily scale to 100 sites in a VPN. The routers
don't all peer
with each other; you (the CE router) only peers with the PE
Router for a
local site. (Each router only has one peering). The
Carriers' routing
takes care of the distribution of your prefixes through
their backbone. It
is (should be) transparent to you. What you have to be wary
of is if the
carrier is using Inter-AS interconnects to extend their
reach.
>
> 3.) The vendor claims they can load balance there own
lines running iBGP
> with another line from another carrier with EIGRP and
no BGP with the
> carrier. I don't think this is possible.
The downside of using BGP and redistribution is the loss of
native routing
metrics and the changed AD of routes (when dealing with
EIGRP, what were
native announcements become Externals after
redistributions.). I've seen
knobs on Cisco equipment that use communities to try to get
around that, but
have never seen it in practice/production.
>
> 4.) This point is kind of selfish, but with all the
buzz about MPLS, will
> this project make me more marketable?
It can't hurt... ;)
>
> Thanks in advance.
>
> MPLS Newbie
Drop me a line offline, and I'll discuss other options in
detail.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7
&i=114087&t=114045
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.
groupstudy.com/list/cisco.html
|