EENY 2010: Packet vs. Optical

8:00 PM -- The subtext to packet/optical convergence is a rivalry between packet vendors and optical-gear vendors. Matthew Finnie, CTO of Interoute Communications Ltd. , proposes a coalition between the two, but that doesn't seem likely.

They're working with opposing motivations. Router vendors preach IP over DWDM, a means of skipping optical transponders. Companies coming from the optical direction say new packet-optical transport systems (P-OTS) can be used to bypass core-router ports.

Both sides say they expect to coexist in peace and harmony, and it's true that neither routers nor optical boxes are in danger of extinction. But it's obvious that companies are approaching packet/optical convergence from two competing angles.

Which brings us back to Finnie. Again. You'll recall he ranted about 100 Gbit/s at last week's Ethernet Expo Americas. (See EENY 2010: Carrier Wants Cheap 100GigE Now Please and EENY 2010: 100G Complaints Continue.)

During one of those panels, he also mentioned his coalition idea, saying packet and optical vendors "need to work together... We're looking for you to drive out something that is cost effective. Otherwise, the whole thing will stop."

Houman Modarres of Alcatel-Lucent (NYSE: ALU) was on the panel, and he had a good point about that word: "I guess you could look at Guantanamo as a coalition, too. People did different crimes and ended up in the same place."

His point with the analogy was that a coalition could end up being an opportunity for everyone to suffer together.

I think Finnie's head and heart are in the right places in suggesting cooperation. He believes it could breed success for the equipment vendors while also saving costs and trouble for his company. And it's not impossible. Juniper Networks Inc. (NYSE: JNPR) and Nokia Networks are working together -- although some, including analyst Ed Zabitsky of ACI Research , thinks NSN has the upper hand in that partnership. (See Is Juniper Junior-Grade?)

But the tendency of every equipment vendor is to protect its own franchise, and that's why I think you'll see the router companies (Cisco Systems Inc. (Nasdaq: CSCO), especially) continue to champion the packet side of packet/optical, while the optical companies will emphasize router bypass.

When I chatted briefly with Finnie, he seemed to realize that his coalition probably won't happen. But in defending his idea during the panel, he pointed out that not every equipment vendor can get their way. "Someone's got to give up something someday. You can't all be king of the hill forever."

— Craig Matsumoto, West Coast Editor, Light Reading

Page 1 / 2   >   >>
spc_markl 12/5/2012 | 4:19:24 PM
re: EENY 2010: Packet vs. Optical


In the very long term, in order to have the flattest, most seamless and lowest cost network, the industry has to move away from IP and do optics all the way through.

Mark Lutkowitz, Telecom Pragmatics

spc_markl 12/5/2012 | 4:19:23 PM
re: EENY 2010: Packet vs. Optical

Just a simplification of architectures in purging protocols geared for electronics would be a welcomed relief.


neyo 12/5/2012 | 4:19:23 PM
re: EENY 2010: Packet vs. Optical Myview,
Do today's packet networks actually do routing? I mean really? Then why do we need mpls? For switching. Right? And as we all know switching is done most optimally at the optical level.
myview 12/5/2012 | 4:19:23 PM
re: EENY 2010: Packet vs. Optical

Interesting statement...and that is based on what?

In which sense is optical better than IP to build a network in the long term? 

Assuming that we left out of the discussion the dream of optical packet switching, which is still a dream, how can you argue that a circuit based network (which is what an optical only network is) is better than a packet based network (which is what IP is about)? This is a very basic and old discussion on the telecom industry, and anyone that states that undoubtfully circuits are better than packets or the other way around, packets are better than circuits, is just wrong. 

Those switching technologies have been created with very different goals and the convenience of one over the other on a specific scenario depends on many factors. Neither IP nor optical are absolutely better than the other, even, and particularly, in terms of cost.

And by the way, you mention, "very long term". I tend to think in the very long term M2M are going to be the predominant source of communications on the networks. What do you think it is better for such scenario, packet switching or circuit switching?





myview 12/5/2012 | 4:19:22 PM
re: EENY 2010: Packet vs. Optical


  "Today´s packet networks" is an extremely broad generalization, and you know it. There are networks where there is need for more routing, and there are networks where there is need for less routing.

  The fact that there is MPLS or not on a network has nothing to do with the need or not to do routing. MPLS is just a switching mechanism. It really depends on how many places on your network you want to take routing decisions. But the network itself, is doing routing anyway. 

  I find it really funny the statement: "And as we all know switching is done most optimally at the optical level". Pls, don´t take my comment aggressively ok? but you and all those that supposedly know that, should go an read some very basics about telecommunications theory and figure out the difference between packet switching and circuit switching. And then analyze more rigorously why and when is circuit switching more optimal than packet switching, and when is the other way around. 

   You can say it loud, many times, in capital letters, as you prefer, but that will never make it true. Optical/circuit switching is not more optimal than MPLS/packet switching, and viceversa, MPLS/packet switching is not more optimal than optical/circuit switching either. Both are different animals applicable to different scenarios. Let´s make all us a favor and remove all this "religion" pushed from vendors. 

  In fact, "packet vs. optical" is in itself a wrong statement, from my perspective, it should read "packet vs. circuit", because isn´t it that the interfaces of the packet based switching nodes are optical as well? what is the essential meaning behind "optical"? Circuit, so let´s be more accurate "packet vs. circuit", however, this title would not be so sexy, I guess.



ethermac 12/5/2012 | 4:19:20 PM
re: EENY 2010: Packet vs. Optical

"Move away from IP"

Sorry, I don't understand this statement. Is that we are going to transport another protocol? Like what, exactly? IPv6? Or a completely new thing?

IP is NOT a routing or switching technology. One needs a routing or switching technology to move around IP datagrams (not packets), and it's better oriented to work on packet switching link layers.

You can work with IP over circuits, but at the end points of these circuits you'd better put something that understand this datagrams and it's able to route (or switch) them.

If you can do this optically (i.e. without OEO conversion) and datagram-by-datagram you've found the holy grial.


joferrei 12/5/2012 | 4:19:19 PM
re: EENY 2010: Packet vs. Optical

We may not, with the event of adaptive 'optics' and generic packet switching technologies such as MPLS-TP, be that far from seeing a break through innovation, by having the optical (speaking loosely--could be any L1/0 media) transport adapt to L3/2 packet layer bandwidth demands.

See eg "Router Driven Lightpaths" concept from slide 8 onward at:


Some qualifiers:

- The adaptive 'optics' should be allowed to supported also on non-L3-router platforms, eg on L2 MPLS switches; the choice of implementation architecture should be up to the network operator.

- The term 'optical' should be understood to cover any and all L1/0 transport mechanisms, be it electrical, optical or even wireless, and could use any T/F/WDM mechanism. As even the above CSCO presentation alludes to, 10Gbps+ transport capacity allocation granularity is overly coarse for packet bursts, and more fine grained capacity allocation adjustment will likely be more efficient.

- In this regard, it is seen that in deed there is nothing magical about optics/DWDM as a media: optical is merely an alternative to electrical and WDM an alternative TDM. These techniques can be used in concert, and in fact the packet layer will not know whether the phys layer muxing and transport was electrical or optical, or used TDM or WDM, or a combination thereof.

In any case, the key to efficiency appears to be making the L1/0 xDM adapt to packet loads, based on desired algorithm, to minimize the cost per bps of packet traffic, while also eliminating most of the transit packet switching/routing points and thus providing high QoS for high revenue applications.

paolo.franzoi 12/5/2012 | 4:19:19 PM
re: EENY 2010: Packet vs. Optical


Simpler solution for Interoute....

Just get every Carrier and every Enterprise to spec in only one product to deliver only a single service.  That would save everyone a lot of time and money.  We can make it work through a single OSS/BSS environment.  Boy this is a great idea!






Duh! 12/5/2012 | 4:19:19 PM
re: EENY 2010: Packet vs. Optical

Yet another case of plus ça change, plus c'est le meme chose.

We've been having these futile circuit vs packet, connection vs connectionless debates for longer than the 30+ years I've been in the industry, and I'm sure we'll still have them when I'm long gone. 

In reality, these constructs are optimized to different points, and need to work together.  Connectionless (or, rather, state-reduced) packet requires no processing or latency for establishing state (connections);  however, forwarding decisions are relatively computationally expensive (longest match vs table indexing), and routing changes take time to propagate and stabilize.  It is for this reason that an underlying mesh of stateful/connection oriented packet (MPLS or ATM) or circuit (OTN or SONET/SDH) is typically needed in the core.  One can argue whether this mesh needs to be more or less dynamic, and how that decision interacts with the routing system.  One can also argue over the degree of the mesh.  There is a case to be made, and rebutted, that long-lived data flows benefit from having their own mesh connection. 

Such are the makings of marketecture. Ultimately operators and enterprises are going to be sold on solutions based on their perceptions of their own needs.  Truth is, no single element or solution is going to ultimately win out.  Hasn't happened over the past 30 years, and no reason to believe the next 30 will be any different.

joferrei 12/5/2012 | 4:19:18 PM
re: EENY 2010: Packet vs. Optical

By innovation, I refer primarily to simplification, ie, ability to achieve the desired end result--in this arena, optimal use of network resources for time-variable bandwidth demands between routers and applications behind them--with minimum overall complexity.

My question here is that, IF routers (as the CSCO presentation implies) are able to manage optical wavelength frequency slot allocations dynamically, why wouldnt they be able to do the same also for the plain digital time slots?

That would allow leaving the optics (with their PhD level analog physical phenomena) static while the packet bursts are accommodated in digital domain, however without individual packet layer processing in the inter router transit network. One could implement such transport switches with off-the-shelf digital logic and standard optical x-ceivers. It appears that the efficiencies can be gained by having the right bandwidth allocation algorithm, whether that algorithm drives (complex) optical frequency slot allocation or (simple) digital time slot allocations.

Naturally this algorithm logic can be included either at router WAN interfaces, or at 'optical' muxes.

Page 1 / 2   >   >>
Sign In