x

The Shorter Roads to 40G

Despite the gloom of the market and the widespread downturn in core optical network spending this year, I remain a believer in 40G. Having spent the past few weeks getting caught up with vendors and carriers on just how the need and applications for 40G are evolving as the market goes through its ugly cyclone, I met no one who said it wasn’t going to happen – and that includes every carrier polled. Obviously, there was a wide range of opinions as to when and how, and herein lay some interesting observations and a few surprises.

The applications for 40G in the network are widespread and not limited to the obvious long-haul DWDM transport network. In fact, there are a number of shorter roads to 40G, and while folks whine about the abysmal core transport market this year, router and switch vendors have been busily defining 40-Gbit/s interfaces that send signals from rack to rack or across town. The surprise in 2002 and 2003 will be just how much the router drives 40G, not the transport network.

In our 40-Gig Forecast last year, we took a detailed look at 40G in DWDM networks and found that, although plenty of technical challenges exist, 40G was, in fact, possible and primarily faced only economic hurdles to deployment. There are two such hurdles:

  • First, line interfaces must be cost competitive with four 10G interfaces and on a steeper cost reduction curve, so that eventually 40G will be only 2.5 times the cost of four 10G interfaces. This, admittedly, will take some time, so initially 40G will be a niche, fitting in where its economics are most compelling.

  • Second, propagation costs must also be factored in, while even if it is cheaper to use 40G line cards it still may be more expensive to carry 40G on a DWDM network because of additional requirements for Chromatic Dispersion and Polarization Mode Dispersion (PMD) compensation and the difficulty involved in adding 40G wavelengths to existing DWDM systems.

The additional costs associated with 40G propagation have clouded the 40G picture and led many observers to count it out for at least three years.

But there are more angles to 40G than capex alone. Opex will drive 40G, and many carriers are beginning to weigh in heavily as interface specifications are being determined, saying they want 40G soon (sometime this year to trial, though expect limited rollouts in 2003), and they want it in more places than just the core DWDM network. Why? Because opex is typically twice the cost of capex, and carriers have become hypersensitive to the effects of opex on their income statements and stock prices this year.

Economic models for incumbent carriers (those still with us), take into consideration buildout and capacity planning scenarios beyond the near term, not simply for a six- or 12-month period. First, they factor in the cost required to go back into a site and add additional capacity or change trunking/interconnect topologies. Simple arguments completely ignore these types of issues. Carriers that have been around a while tend to have interconnect strategies with two- to four-year planning horizons. Additionally, numerous generations of switching and routing equipment have shown that maximum performance metrics (bandwidth, power, etc.) per rack unit are achieved with increasing port speeds. First-generation development efforts do not necessarily prove this in, but second generation always has.

Like many new technological steps, 40G will likely follow a path of least resistance into the network. These “resistances” can be classified as either technological or economic and can weigh equally heavily on a technology as it makes its way to market.

Here’s a look at how 40G is arising:

  • Ultra Long Haul (600km and up): Here, where distances exceed 600 kilometers, 40G runs into serious propagation issues, though many of the transponder vendors I visited were able to demonstrate 40G at distances of at least 1,000 km over standard SMF or NZ-DSF fibers currently deployed in major carrier networks. Breaking through the 1,500km barrier is going to be tough for 40G, and carriers will have to decide whether they can accept network architectures in which the longest route is 1,500 km. Some are already asking for more, especially those envisioning photonic switching at core nodes, which adds loss – further limiting the reach of 40G. The worry here is that as 10G transport systems reach 3000 km, how will carriers mix and match 40G on those systems if distance limitations are quite different? This could cloud the opex savings if a carrier is forced to make sub-optimal network architecture decisions based on 40G's differing propogation constraints.

  • Long Haul (around 500-600km between regeneration): As discussed above, this is a market that is right for 40G, but the current state of the market is depressed to the point that many carriers are putting off adding new routes or overbuilding existing routes with new equipment, so 40G may find it can only enter the market in span-by-span upgrades, where capacity is choked. This puts significant pressure on vendors to make sure 40G can be added to systems designed around 10G – so issues of dispersion compensation and nonlinear effects can be dealt with at the transponder, rather than out in the network, where forklift upgrades would be required.

    40G actually brings a rather unique proposition to the table: In the past, service providers were forced to make fundamental changes in their networks to increase capacity: introduction of in-line Optical Amplification, the concept of Wavelength Division Multiplexing (WDM), as well as the transition from 2.5 to 10 Gbit/s, which introduced the concept of in-line dispersion compensation. All of these developments required forklift upgrades of the installed system as well as a new network planning approach. The upgrade to 40 Gbit/s will in most cases be trivial in comparison, only involving changing the transponders at the terminal sites – as long as those transponders have the necessary intelligence to adapt to the propagation issues arising out of adding 40G to a 10G line system.

  • Metro DWDM (up to 150 km): This is perhaps the path of most resistance, since the demand pull is not very strong, the distances are long enough to require both chromatic and polarization mode dispersion compensation, and cost pressures are highest. Thus, 40G will have to wait the longest to appear here, maybe two years, and even then it will represent a niche, as 10G does today. This will be the most hotly debated area for 40G, since in many carrier networks OC48 (2.5 Gbit/s) continues to rule the day. Making the transition to 10G has been difficult in RBOC networks because of their aging base of embedded fiber, so 40G may be out of the question in those networks. For IXCs and out-of-region ILECs the story is different, however, and for those carriers that are building out high-capacity metro networks that are designed primarily to carry data traffic, the economics may soon favor 40G. Look for some trials this year and some limited rollouts later in 2003 for specialized applications.

  • Intermediate Reach/Long Reach (40-80 km): In this application, 40G will not always be implemented as part of a DWDM system but will often be direct to fiber; it will appear both on routers and some transmission equipment, such as bandwidth management optical switching systems. The router implementation seems more likely in the near term (18 months) because a number of data network operators see the value in pushing content closer to end users, and they can therefore take advantage of the economics of 40G router interfaces with these reaches to interconnect POPs (points of presence) without getting on the transport network at all.

    As the metro network develops into more than just an area of aggregation, metro-centric data and content applications are emerging that will reinforce the need for intermediate routing or packet switching. As these applications become real-time, the benefits of higher-speed serial interfaces (stream continuity, sequencing, etc.) become more obvious. This is already appearing at OC192 and 10-Gigabit Ethernet, so it should follow that 40G will be similarly adopted. The opto-electronics on these 40G interfaces are rather straightforward; the major development challenges are the packet processing functions (lookup, classification, IP forwarding, and queuing) and maintaining packet order in a 40G stream.

  • Short Reach (2-40 km): Short-reach interfaces for 40G may reside on any variety of transmission equipment or data equipment and provide very fat pipe connectivity among a carrier’s facilities. The applications for this are rather broad and include optical switching systems, which will eventually interface directly to routers and to long-haul DWDM systems at 40G. The optical switch is emerging as a better mux than an ADM because of its improved economics and its intelligence, so ultimately the edge/groomer optical switch will be in charge of the optical handoff to long-haul. With 40G short-reach interfaces, an optical switch can provide handoff to long-haul gear that isn’t collocated. Additionally, optical switches or routers can take advantage of short-reach 40G to extend the physical dimensions of a network element, creating a “logical” optical switch or router out of shelves in different carrier POPs.

  • Very Short Reach (VSR) (up to 2 km): These interfaces are designed to provide low-cost interconnection of collocated equipment, such as routers, optical switches, and Sonet (Synchronous Optical NETwork) and SDH (Synchronous Digital Hierarchy) ADMs. At a POP or central office, connections into the facility – and within the facility – all terminate on the patch panel and/or optical switch (OXC). Typically, there are no direct connections among network elements in a POP. Crossconnects are made at the fiber patch panel or, in cases where one exists, the optical crossconnect. Considering that equipment may or may not be in the same room or floor, the distances and types of connection are significant concerns, especially when running at these very high speeds or across purely optical systems that introduce loss. Measuring distances between boxes is time consuming and prone to human error, and using multiple cables or ribbon connections is operationally prohibitive because it increases cost per network connect. Adding protection to each connection further complicates the situation.

The Optical Internetworking Forum (OIF) has been working on 40G as part of their specifications for VSR interfaces and is in line with the International Telecommunication Union (ITU)’s standardization process for VSR (G.693). The VSR-5, as the OIF calls it, will be suitable for an SFI-5 electrical interface carrying OC768 Sonet framed data.

People tend to think of VSR as the province of parallel optics – 40G implemented as an array of lower-speed Vertical Cavity Surface Emitting Lasers (VCSELs) and ribbon fiber or, in some cases, as four 10G streams using coarse WDM. This will certainly be the case in many instances, but even at 10G, serial VSR has important advantages over parallel optics, and these will likely obtain at 40G, as well. These include leveraging the installed base of singlemode fiber within an office rather than moving to ribbon fiber for VCSEL arrays; test equipment considerations, serial being easier than parallel; and fewer personnel and training procedures for dealing with parallel VSR, as everyone understands serial. Finally, serial goes farther than parallel, as a rule, and in many offices equipment resides on different floors, not just in neighboring racks, or traverses numerous patch panels or all-optical switch fabrics, favoring the longer-reach of the 40G serial.

It’s interesting to note how well the 40G food chain is being developed, despite the broad perception that 40G is lying in the same grave as bankrupt backbone carriers. At the Optical Fiber Communication Conference and Exhibit (OFC) this year, 40G will again be a hot topic, with many announcements from public and private companies delivering all the necessary electronics, opto-electronics, and optics to support 40G, both long and short haul, for the data network and the transport network. Here’s a quick list of the latest developments:



What are the implications of all this activity? During my recent tour of companies I found a real opportunity has emerged for 40-Gbit/s transponder and subsystem suppliers, since many of those projects within the big vendors have gotten cut during recent downsizings. This has led to a continuing need for solutions at 40G, but delivered now as packaged modules or subsystems, rather than piece parts to be engineered into new systems. This trend will likely be common not just in the 40G space, but also in other areas such as photonic switching, optical add/drop, gain management, and monitoring. Looking for bright spots in this market is hard work, but 40G, I’m still convinced, is one of them.

— Scott Clavenna, Director of Research, Light Reading
http://www.lightreading.com
Page 1 / 3   >   >>
dwdm2 12/4/2012 | 10:47:26 PM
re: The Shorter Roads to 40G "The Shorter Roads to 40G" a wonderful report by Scott Clavenna.

It appears that, the biggest challenge is NOT the technology for dispersion control but the market.

While 40G is a must for meeting the bandwidth demand, the vendors who are actually in the market, are still sorting out OC-48 and trying to get into OC-192.

So whatever activities going on around OC-768, are still based on future predictions, with a hope that the investment will pay off!

Regards,
AR
optigirl 12/4/2012 | 10:47:25 PM
re: The Shorter Roads to 40G I have heard mixed stories about the potential for this market but I would tend to agree that we will someday live in a world where 40G is real, deployed and profitable, the latter being the biggest issue of all.

Who will buy this stuff? Not the RBOCs, and I doubt that Sprint and Qwest will be doing it anytime in the next 3 years. AT&T still has a lot of old fiber laying about so that would be an issue for them as well. I would bet on a 5 year plan for something significant to happen beyond the normal vendor hype, sponsored consulting plugs or one-off situations. How long did it take for us to get from OC-3 to OC-12 to OC-48 to OC-192.

A lot needs to happen before we get to 192 and while the advocacy for the technology is certainly something nice, I think you might be getting excited about possibility and potential instead of what history and our experiences have shown us to be true.
skeptic 12/4/2012 | 10:47:24 PM
re: The Shorter Roads to 40G Additionally, optical switches or routers can take advantage of short-reach 40G to extend the physical dimensions of a network element, creating a GǣlogicalGǥ optical switch or router out of shelves in different carrier POPs.
--------------------

Most people will not do this. The problem with
"logical" routers is that they are terribly
inefficient:

if you connect two elements:

- You have to buy two OC-768 ports

- waste one or more slots on each device

- Manage and monitor two elements

Now if you connect four elements, the number
of slots/ports wasted goes up dramatically.
The only way to keep it from going up is to
compromise on bandwidth between elements.

Fewer network elements is always better.

Every extra interconnect within the central
office is a waste of money and increases
operations cost.


dwdm2 12/4/2012 | 10:47:24 PM
re: The Shorter Roads to 40G Who will buy this stuff?

Actually the big gyus have already started probing into the gadgets necessary for 768. Because the technology is still not matured yet, and even for the alpha or beta (that are available) the cost is prohibitive for small guys.

This is where the vacuum is. Its the small and medium companies and the service providers who will make the market vibrant...but they are completely out of touch!

AR
optigirl 12/4/2012 | 10:47:24 PM
re: The Shorter Roads to 40G A lot needs to happen before we get to 192 .....

Correction: 768.

Need more coffee I guess....

:)
melao 12/4/2012 | 10:47:23 PM
re: The Shorter Roads to 40G "Who will buy this stuff?

Actually the big gyus have already started probing into the gadgets necessary for 768. Because the technology is still not matured yet, and even for the alpha or beta (that are available) the cost is prohibitive for small guys.

This is where the vacuum is. Its the small and medium companies and the service providers who will make the market vibrant...but they are completely out of touch!"

I have another view also. My guess is that 40G will be mostly on the core. And i see High Channel Count DWDM systems to pop-up more and more. Because without that the whole philosophy of all-optical network wouldn-t work. Which is driving lot-s of researches from a lot of vendors.

Like, nowadays NT claims to have completed a system with 160 wavelengths, 10Gig each one. Well, i looked on papers of that project (the terabit challenge). Let me tell you it was only possible because the line-amp sites were luckly placed correctly. It won-t be possible in 90% of the networks.
So if with 10Gig the Hi Channel Count systems have a LOT of trouble, imagine 40Gig.

I work on a carrier, the only link that we have 32 wavelengths uses 2.5G, because it wouldn-t be possible with 10Gig.

Well, that-s it! Cheers!
Edge0fSpace 12/4/2012 | 10:47:23 PM
re: The Shorter Roads to 40G Actually, most of the "next-gen" (I use this term somewhat lightly, and loosely) router vendors out there are looking into ways to connect the fabric together. (Not line cards.) This makes logical routers considerably more efficient than the scenario given. (All your line cards are revenue producing.)

E0S
dwdm2 12/4/2012 | 10:47:23 PM
re: The Shorter Roads to 40G Every extra interconnect within the central
office is a waste of money and increases
operations cost.
-----------------------

And I'll add: Every extra interconnect adds several degrees of network complicacies and may end up with fewer available channels.

However, while this scenario appears to be discouraging for 40G, I hear few companies are thinking of an "Integrated solution" where all the boxes will be integrated into one!

Does anyone know more specifics about such efforts?

AR
PhotonGolf 12/4/2012 | 10:47:22 PM
re: The Shorter Roads to 40G
First, I have to say that I'm biased. OK, got that out ...

I see the guts if the 40G transponder already approaching the costs of the 10G guts just two years ago. And these prices are falling constantly. There is new technology just coming available which will drive this curve down, and, provided greedy vendors aren't too greedy, I believe 40G transponders at the 2.5x10G price point will be here easily in 2003.

If that's the key as Scott says, lots of things line up.

P

PS I'm really talking about LH and discreet transponders rather than some fantasy integrated solution.
dwdm2 12/4/2012 | 10:47:22 PM
re: The Shorter Roads to 40G I work on a carrier, the only link that we have 32 wavelengths uses 2.5G, because it wouldn-t be possible with 10Gig.
-----------------------

My hunch is this owing to the legacy switching technology and godd old TFF based dwidum! I guess you could use an upgrade to 40G! :-)
Page 1 / 3   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE