[190583] in North American Network Operators' Group
Re: L3 over OTN vs pure IP/MPLS
daemon@ATHENA.MIT.EDU (Mark Tinka)
Sat Jul 9 05:23:39 2016
X-Original-To: nanog@nanog.org
To: =?UTF-8?Q?Manuel_Mar=c3=adn?= <mmg@transtelco.net>, NANOG <nanog@nanog.org>
From: Mark Tinka <mark.tinka@seacom.mu>
Date: Sat, 9 Jul 2016 11:23:29 +0200
In-Reply-To: <CAD0TWZ_kUTO2utNddusizoOngHUP9MVesyqSXpkPu8A1GqXSSg@mail.gmail.com>
Errors-To: nanog-bounces@nanog.org
On 8/Jul/16 04:43, Manuel Mar=C3=ADn wrote:
> Dear Nanog community
>
> We are currently experimenting TCP degradation issues in some metro mar=
kets
> where there are multiple POPs and the IP packets have to pass multiple =
L3
> access devices (routers) before reaching the core router. The more L3 h=
ops
> that it goes through the more degradation we see in the Internet servic=
e.
> L3 routers have the capacity in terms of pps and BW and we are using ju=
st a
> small percentage of that capacity but anyway the service is degraded (N=
o
> CRC/Input errors in the links). A couple of vendors are recommending us=
ing
> OTN or MPLS-TP to backhaul the connections from the access to the core =
but
> obviously deploying OTN means a considerable amount of capex. The
> flexibility and simplicity of the L3 based access network is great but
> given the performance issues we are experimenting I would appreciate if=
you
> can share your experience/recommendation with either OTN or MPLS-TP. Do=
es
> it really make sense to use a OTN/Photonic layer to backhaul the L3
> connections in a metro network?
Are you running your Metro-E networks on dark fibre all the way into
your core, or are you leasing a lit service from another carrier for this=
?
If the former, you should not have any issues with performance
considering a well-designed network. If the latter, you are in the hands
of your carrier, and you would need to qualify that they know what they
are doing before you overlay your network on top of theirs.
I cannot find any benefit in running your Metro-E network on dark fibre
vs. OTN into the core apart from when you hit the port bandwidth limits
in your Metro-E network and need to go to DWDM to upgrade that so as to
keep the fibre-count down while increasing bandwidth.
We run a large dark-fibre-based Metro-E network without any DWDM
equipment between most rings and our core, and there is no tangible
difference in performance between services delivered on those rings or
services delivered upstream in the core data centres. There is also no
difference in performance between the dark fibre rings vs. the DWDM
rings (although, depending on your equipment vendor, some platforms
could introduce quite a bit of serialization delay, but not enough to
affect actual performance).
Sounds like your vendors are trying to sell you boxes...
Mark.
