[c-nsp] MPLS-TP on CPT platform vs IP/MPLS core on ASR with TE

[Phil Bedard]

On 11/27/13, 9:47 PM, "Mark Tinka" <mark.tinka at seacom.mu> wrote:

>On Wednesday, November 27, 2013 09:50:16 PM Phil Bedard
>wrote:
>
>
>
>> Cisco seems to have abandoned the concept of putting the
>> optics on the router, because the density isn't there.
>> That's always been the rub with colored/tunable optics
>> when you are on the bleeding edge of bandwidth, the
>> density isn't there.  There was no way they could do
>> 10x100G on the NCS6K using tunable optics.  They are
>> more inclined now on the NCS6K and ASR9K to always pair
>> the router with a transport shelf and then use
>> control/management plane magic to make them look like
>> one virtual entity.
>
>This has been their model since they began their IPoDWDM
>schpill on the XR 12000 platform.

It hasn't really up until recently.  We still have some of the CRS 40G
StrataLight IPoDWDM modules... And previously they were pushing the notion
of a "virtual transponder" on the router controlled by CTC, using a
tunable optic on the router itself.  They even built a 1x100G DWDM card
for the CRS. With the CRS-X and the ASR9K they introduced the nV-Optical
setup of having a remote DWDM shelf under one management/control plane.
Really both methods still fall under the "nLight" and "iOverlay" paradigms
and they still support both and talk about them.  Really 100G is what has
pushed the new concepts, and the evolution are things like variable rate
ethernet with a 1:1 mapping to superchannels on the DWDM line side.  So if
you need 350Gbps between two sites you can get it on a single interface.
But that's a couple generations of hardware away and at least on the DWDM
side completely non-interoperable. :(

>
>> CPAK gives them the ability to use a
>> very low cost interconnect, or they can use traditional
>> transponders.  It works very well for 100G since you can
>> have a 1:1 relationship between router egress port and
>> line-side wavelength, treating them as a single entity.
>
>> However that requires you buy into the "NCS" as a
>> complete system, or at least use a Cisco router
>> (NCS/ASR9K) along with their transport shelf. I know
>> they are working to standardize the iOverlay stuff but
>> you know how well this stuff interoperates...
>
>And this is why I still have my doubts about the practical
>benefits that IPoDWDM will bring to the majority of
>operators, particularly those that don't run their own
>optical backbones.
>
>If there is still a physical difference between the optical
>and routing gear, the choice of whether you use colored or
>grey light becomes an obvious decision for most operators.

Yes, you definitely need to run your own fiber network to make real use of
it.  


>
>> Juniper coming out with a DWDM PIC with the same density
>> is a pretty big feat, even if the PIC weighs like 40lbs.
>> :)  However, they will soon double the density on the
>> PTX most likely and leave the DWDM solution at half the
>> density.  Some transceiver vendors developed a OTU4 DWDM
>> CFP but they haven't seen the light of day, however
>> tunable OTU4 CFP2 has already been shown.  So the future
>> for Juniper and other vendors is to use tunable
>> pluggable transceivers instead of built-in optics.
>> Cisco with the CPAK will initially be shut out of that I
>> imagine.
>
>The CPAK is a pretty good - what they have done there with
>CMOS photonics is not to be taken lightly. My expectation is
>that if they would like to have a tunable CPAK module, they
>can develop that (but probably won't, as it affects their
>IPoDWDMv2 campaign - NCS2000 + NCS6000, e.t.c.).
>
>Personally, I don't mind whether the MSA standardize the
>CPAK or not, in the short-to-medium term. 100Gbps Ethernet
>purchases are quite specific that it is likely to still make
>sense to buy 100Gbps ports from Cisco together with the
>optics, and beat them enough to sell you the CPAK for close
>to nothing, mostly because of the number of ports you can
>get into the line card, and the heat and power savings that
>CPAK offer today.
>
>Naturally, in the long term, I hope the industry can have a
>reasonably small pluggable that all vendors agree on.


Yeah it's called the MSA CFP2/CFP4 :).  The CFP2 uses marginally more
power than the CPAK (6.5-7w vs. 5.5w) and generally can fit just as many
on the same size faceplate.  They will also support some of the fancy CPAK
stuff like a 10x10 breakout.  Cisco was ahead in the space/power debate
but not really anymore.  There are a bunch of vendors making CFP2 and
you'll see modules come out next year from most vendors.  CPAK does give
Cisco a lot of flexbility in the realm of cheap custom transceivers,
similar to what they did with the Nexus FEX transceivers.  If they can
make the router/transport interconnect $0 it's a big win.   But in the end
I see it as X2 v2 and will fade over time.  Finisar has already
demonstrated a working CFP4 module which is half the size of the CPAK and
uses about 3.5w.  


>
>Cheers,
>
>Mark.


Phil