Worldwide Ethernet switch and Router Market on the Wing

The worldwide Ethernet switch market (Layer 2/3) reached US$5.66 billion in the third quarter of 2013, representing an on-year increase of 6.5%. On a sequential basis the third quarter of 2013 increased 5.2%. Meanwhile, the worldwide router market jumped 7.8% on year, which was the highest annual growth rate seen in over two years. Sequentially, the router market increased 2.2%.

From a geographic perspective, third-quarter results show the Ethernet  switch market had its highest growth in Middle East and Africa (22.7% on year) and very good growth in North America (12% on year). Latin America and Asia Pacific (excluding Japan) both grew in the mid-single digits while the Western Europe and Central and Eastern Europe regions were up in the 1-2% range. Meanwhile, Japan declined 13.4% on year in the third quarter of 2013.

10Gb Ethernet  switch (Layer 2/3) revenues increased 12% on-year while 10Gb Ethernet switch port shipments grew 40.1% on year to 4.97 million ports in the third quarter of 2013. 10GbE continues to be the primary driver of the overall Ethernet switch market, with the early stage 40GbE just starting to reach critical mass as a stand-alone segment.

The worldwide enterprise and service provider router market also performed well in the third quarter of 2013, increasing 7.8% on year and 2.2% sequentially in the quarter. Regionally, results were quite uneven, ranging from more than 30% growth in Middle East and Africa and Central and Eastern Europe to a significant 26.3% decline in Japan. North America increased 14.8% on year while Western Europe and Latin America were up high single digits. Asia Pacific (excluding Japan) declined 1.5% on year in the third quarter of 2013.

From a vendor perspective, Cisco's Ethernet  switch (Layer 2/3) market share in the third quarter of 2013 held steady from previous quarter at 62.3%. Cisco's market share in the fast-growing 10GbE market segment was 65.9% in the third quarter of 2013. HP's Ethernet switch revenues increased 5.3% on quarter and 6.5% on year in the third quarter of 2013 on par with the overall market. HP's market share now stands at 9.4% in the third quarter of 2013.

With the unified enterprise network edge gaining market traction and new applications for the Internet of Things continuing to emerge, Gigabit Ethernet (GbE) is expected to hold its own for the foreseeable future. Of course, noticeable regional and country-level variances continue to be seen, so vendors are best advised to take those into consideration while working on their near to mid-term strategies.

IDC: Ethernet  switch revenues by vendor, 3Q2012-3Q2013 (US$m)
	3Q 2012	4Q 2012	1Q 2013	2Q 2013	3Q 2013
Cisco	3328.85	3506.59	3310.39	3365.28	3543.95
HP	498.31	532.76	459.8	507.91	531.77
Alcatel-Lucent	167	172.8	185.03	199.42	204.36
Juniper	142.86	144.22	130.49	160.77	150.63
Dell	96.94	121.73	124.13	96.18	111.14
Others	1080.64	1209.79	1023.94	1055.33	1120.4

Network Crunch is Coming, How to Tackle?

                        

A European Union research project is looking at ways to expand by one-hundredfold the capacity of an optical fibre. The project will need to develop techniques to exploit multimode and specialist fibre, and if successful, will require new fibre deployments in the network.

Current 100 Gigabit-per-second (Gbps) dense wavelength division multiplexing (DWDM) systems are able to transmit a total of 10 terabits-per-second of data across a fibre (100 channels, each at 100Gbps). System vendors have said that with further technology development, 25Tbps will be transported across fibre.

But IP traffic in the network is growing at over 30% each year. And while techniques are helping to improve overall transmission, the rate of progress is slowing down. A view is growing in the industry that without some radical technological breakthrough, new transmission media will be needed in the next two decades to avoid an inevitable capacity bottleneck.

"The Shannon Limit - the amount of information that can be transmitted - depends on the signal-to-noise and the amount of power you can put down a fibre," says Ian Giles, CEO of Phoenix Photonics, a fibre component specialist and one of the companies taking part in the project. "You can enhance transmission capacity by modulation techniques to increase bit rate, WDM and polarisation multiplexing but in the end you run out of the ability to transmit more information along a single-mode fibre."

This 'network crunch' is what the MODE-GAP project is looking to tackle.


Project Details

One of the approaches that will be investigated is exploiting the multiple paths light travels down a multimode fibre to enable the parallel transmission of more than one channel.

These multiple paths light takes traveling in a multimode fibre disperses the signal. "The proposal we are making is that we take a low-moded fibre and select specific modes for each channel, or a high-moded fibre and select modal groups that are very similar," says Giles. The idea is that by identifying such modes in the multimode fibre, the dispersion for each mode or model groups will be limited.
But implementing such a spatially modulated system is tricky as the modes need to be identified and then have light launched into them. In turn, the modes must be kept apart along the fibre's span.
The project will tackle these challenges as well as use digital signal processing at the output to separate the transmitted channels. The project consortium believes that up to 10 channels could be used per fibre.

Photonic bandgap fibre has an air core that doesn't create scattering. As a result the non-linear threshold is some 100x higher, meaning that more power can be put into the fibre.


What Next?

The MODE-GAP project is still in its infancy. The goal is to develop a system that allows the multiplexing and demultiplexing of the spatially-separated channels on the fibre. That will be done using multimode fibre but Giles stresses that it could eventually be done using photonic bandgap fibre. "You then enhance capacity: you increase the number of channels, and decrease the non-linearities which means you can increase the amount of information sent per channel," says Giles.
"Up till the spatial modulation part, the system is the same as you have now," he adds. "It is only the spatial modulation part that needs new components." NSN will use any prototype developed within its test-bed where it will be trailed. "They don't want to reinvent their equipment at each end," says Giles.

The project will also look to develop a fibre-amplifier that will boost all the fibre's spatial separated channels.

The project's goal is to demonstrate a working system. "The ultimate is to show the hundredfold improvement," says Giles. "We will do that with multiple channel transmission along a single photonic bandgap fibre and higher capacity [data transmission] per channel."


Project Partners

In addition to NSN's systems expertise and test-bed, Eblana Photonics will be developing lasers for the project while Phoenix will address the passive components needed to launch and detect specific modes. OFS Fitel is providing the fibre expertise, while the University of Southampton's Optoelectronics Research Centre is leading the project. 

The other universities include the COBRA Institute at the Technische Universiteit Eindhoven which has expertise in the processing and transmission of spatial division multiplexed signals, while the Tyndall National Institute of University College Cork is providing system expertise, detectors, transmitters and some of the passive optics and planar waveguide work.
 

How Flexible Grid WDM Evolved?

  

WDM networks operate by transmitting multiple wavelengths, or channels, over a fiber simultaneously. Each channel is assigned a slightly different wavelength, preventing interference between channels. Modern DWDM networks typically support 88 channels, with each channel spaced at 50 GHz, as defined by industry standard ITU G.694. Each channel is an independent wavelength.

The fixed 50-GHz grid pattern has served WDM networks and the industry well for many, many years. It helps carriers easily plan their services, capacity, and available spare capacity across their WDM systems. In addition, the technology used to add and drop channels on a ROADM network is based on arrayed-waveguide-grating (AWG) mux/demux technology, a simple and relatively low-cost technique particularly well suited to networks based on 50-GHz grid patterns.
  WDM networks currently support optical rates of 10G, 40G, and 100G per wavelength (with the occasional 2.5G still popping up), all of which fit within existing 50-GHz channels. In the future, higher-speed 400-Gbps and 1-Tbps optical rates will be deployed over optical networks. These interfaces beyond 100G require larger channel sizes than used on current WDM networks. The transition to these higher optical rates is leading to the adoption of a new, flexible grid pattern capable of supporting 100G, 400G, and 1T wavelengths.

Current generation

The fixed 50-GHz grid pattern specified by ITU standards is shown in Figure 1. Any 10G, 40G, or 100G optical service can be carried over any of the 50-GHz channels, which enables carriers to mix and match service rates and channels as needed on their networks.
A look inside each channel reveals some interesting differences between the optical rates and resulting efficiency of the optical channel (see Figure 1). A 10G optical signal easily fits within the 50-GHz-channel size, using about half the available spectrum. The remaining space within the 50-GHz channel is unused and unavailable. Meanwhile, the 40G and 100G signals use almost the entire 50-GHz spectrum.

FIGURE 1. Optical rates and their spectral efficiency.

Spectral efficiency is one measure of how effectively or efficiently a fiber network transmits information and is calculated as the number of bits transmitted per Hz of optical spectrum. With 10G wavelengths the spectral efficiency is only 0.2 bit/Hz, while the 100G wavelength provides a 10X improvement in spectral efficiency to 2 bits/Hz. The more bits that can be transmitted per channel, the greater the improvement in spectral efficiency and increase in overall network capacity and the lower the cost per bit of optical transport.
While 100G wavelengths are becoming more common, carriers are already planning for higher-speed 400G and 1T channels on their future ROADM networks, with the expectation that spectral efficiency will at least remain the same, if not improve. New ways of allocating bandwidth will be needed to meet these expectations.

Superchannels

As mentioned, WDM networks currently transmit each 10G, 40G, and 100G optical signal as a single optical carrier that fits within a standard 50-GHz channel. At higher data rates, including 400G and 1T, the signals will be transmitted over multiple subcarrier channels (see Figure 2). The group of subcarrier wavelengths is commonly referred to as a "superchannel." Although composed of individual subcarriers, each 400G superchannel is provisioned, transmitted, and switched across the network as a single entity or block.

FIGURE 2. 400G modulation options and superchannels.

While 400G standards are still in preliminary definition stage, two modulation techniques are emerging as the most likely candidates: dual polarization quadrature phase-shift keying (DP-QPSK) using four subcarriers and DP-16 quadrature amplitude modulation (QAM) with two subcarriers. Due to the differences in optical signal-to-noise-ratio requirements, each modulation type is optimized for different network applications. The 4×100G DP-QPSK approach is better suited to long-haul networks because of its superior optical reach, while the 2×200G DP-16QAM method is ideal for metro distances.
Since 400G signals are treated as a single superchannel or block, the 400G signals shown in Figure 3 require 150-GHz- and 75-GHz- channel sizes, respectively. It's this transition to higher data rates that leads to the requirement for and adoption of new flexible grid channel assignments to accommodate mixed 100G, 400G, and 1T networks (see Figure 3).
A new flexible grid pattern has been defined and adopted by ITU G694.1. While commonly referred to as "gridless" channel spacing, in reality the newly defined flexible channel plan is actually based on a 12.5-GHz grid pattern. The new standard supports mixed channel sizes, in increments of n×12.5 GHz and easily accommodates existing 100G services (4×12.5 GHz = 50 GHz) and future 400G (12×12.5 GHz) and 1T optical rates.

FIGURE 3. Flexible grid channel plan.

One of the advantages of the flexible grid pattern is the improvement in spectral efficiency enabled by more closely matching the channel size with the signals being transported and by improved filtering that allows the subcarriers to be more closely squeezed together. As shown in Figure 5, four 100G subcarriers have been squeezed into 150-GHz spacing, as opposed to the 200 GHz (4×50 GHz) required if the subcarriers were transported as independent 50-GHz channels. The net effect of the flexible channel plan and closer subcarrier spacing is an improvement in network capacity of up to 30%.
One common "myth" in the industry is that legacy networks must be upgraded, or "flexible grid-ready," to support 400G optical rates and superchannels. While having flexible grid-capable ROADMs can improve spectral efficiency, they're not a requirement to support 400G or superchannels on a network. Since the subcarriers are fully tunable to any wavelength, they can simply be tuned to the existing 50-GHz grid pattern, allowing full backward compatibility with existing ROADM networks.

FIGURE 4. Transmitting 400G on legacy WDM networks.

CDC ROADMs

Closely associated with flexible grid channel spacing are colorless/directionless/gridless (CDG) and colorless/directionless/contentionless/gridless (CDCG) ROADM architectures. Along with gridless channel spacing, CDC ROADMs enable a great deal of flexibility at the optical layer.
Recall that existing ROADM systems are based on fixed 50-GHz-channel spacing and AWG mux/demux technology. The mux/demux combines and separates individual wavelengths into different physical input and output ports. While the transponder and muxponder themselves are full-band tunable and can be provisioned to any transmit wavelength, they must be connected to a specific port on the mux/demux unit. A transponder connected to the west mux/demux only supports services connected in the west direction. To reassign wavelengths – either to new channels or to reroute them to a different direction – requires technician involvement to physically unplug the transponder from one port on the mux/demux and plug it into a different physical mux/demux port.
CDC ROADMs enable much greater flexibility at the optical layer. The transponders may be connected to any add/drop port and can be routed to any degree or direction. Wavelength reassignment or rerouting can be implemented automatically from a network management system, or based on a network fault, without the need for manual technician involvement. The tradeoffs with CDC ROADMs are more complex architectures and costs.

Flexible Network Muscles

The existing 50-GHz-channel plan based on ITU G.694 has served the industry well for many years. But as the industry plans for the introduction of even faster 400G, and eventually 1T, optical interfaces, there's a need to adopt larger channel sizes and a more flexible WDM  spacing plan.
These higher-speed optical interfaces rely on a new technique involving superchannels that comprise multiple subcarrier wavelengths. These subcarriers are provisioned, transported, and switched across a network as a single block or entity. Flexible grid systems enable the larger channel sizes required by 400G and 1T interfaces, but also allow the channel size to be closely matched to the signal being transported to optimize spectral efficiency.
No discussion of gridless ROADMs would be complete without including new next generation CDC ROADM architectures. These new ROADMs will enable a great deal more flexibility and efficiency at the optical layer.
 

More bandwidth means more testing?

The use of MPO cables for trunking 10-Gbps connections in the data center has steadily risen over the past 10 years. That trunking requires use of a cassette at the end of the MPO cables designed to accommodate legacy equipment connections. Now that 40-Gbps and 100-Gbps connections are coming on the market, a migration path has emerged: Remove the 10-Gbps cassette from the MPO cable and replace it with a bulkhead accommodating a 40-Gbps connection. Then it might be possible to remove that bulkhead and do a direct MPO connection for 100 Gbps at a later date.
The problem is that while this migration strategy is an efficient way to leverage the existing cabling, in comparison to 10-Gbps connections, the 40-Gbps and 100-Gbps standards call for different optical technology (parallel optics) and tighter loss parameters.

In short, each time you migrate you need to verify the links to ensure the performance delivery the organization requires.

To understand the challenges ofMPO cables validation, it’s necessary to understand MPO cables and how they’re tested in the field. An MPO connection is about the size of a fingernail and contains 12 optical fibers, each less than the diameter of a human hair – and each one needs to be tested separately. That traditionally means the use of a fan-out cord to isolate each fiber, followed by tedious manual testing, tracing, and error-prone calculations.

The actual fiber test is quick enough: typically under 10 seconds per fiber once you’re in process. But you better be cruising: While one of our enterprise customers has data centers with as little as 24 MPO cables  trunks (x12 fibers each), that same customer also has a 30,000-MPO data center installation. That’s 30,000 connections with 12 fibers each, or roughly 3,120 hours in labor (and $343,200 in cost) if you had to test them all individually.

And at some point, you better have tested them. There were two primary drivers behind development of MPO cables trunks. The first was the ever-increasing need for cabling density in the data center. Cabling blocks airflow, so the denser the cable, the better the thermal management. And, as data center bandwidth steadily climbs to 10, 40, and 100Gbps, a dense multi-fiber cable becomes the only option.
But the second, perhaps more important factor, is the difficult and highly technical nature of field termination for fiber. We’re talking curing ovens, adhesives, microscopic fibers, etc. Given that expensive and time-consuming “craft” process, modular factory-terminated MPO cables promise simplicity, lower cost, and true plug-and-play fiber connectivity.

The challenge is that pre-terminated fiber is only guaranteed “good” as it exists in the manufacturer’s factory. It must then be transported, stored, and later bent and pulled during installation in the data center. All kinds of performance uncertainties are introduced before fiber cables are deployed. Proper testing of pre-terminated cables after installation is the only way to guarantee performance in a live application. In short, investing in factory-terminated fiber trunks to save time and decrease labor costs doesn’t really offer an advantage if the testing becomes an expensive bottleneck.
Testing and determining fiber polarity is another challenge. The simple purpose of any polarity scheme is to provide a continuous connection from the link’s transmitter to the link’s receiver. For array connectors, TIA-568-C.0 defines three methods to accomplish this: Methods A, B, and C. Deployment mistakes are common because these methods require a combination of patch cords with different polarity types.


You can buy fiber optic jumpers with any connectors from Optoroute.

How to test MPO/MTP Cable?

To understand the challenges of MPO cable validation, it’s necessary to understand MPO cableand how they’re tested in the field. An MPO connection is about the size of a fingernail and contains 12 optical fibers, each less than the diameter of a human hair – and each one needs to be tested separately. That traditionally means the use of a fan-out cord to isolate each fiber, followed by tedious manual testing, tracing, and error-prone calculations.

Testing and determining fiber polarity is another challenge. The simple purpose of any polarity scheme is to provide a continuous connection from the link’s transmitter to the link’s receiver. For array connectors, TIA-568-C.0 defines three methods to accomplish this: Methods A, B, and C. Deployment mistakes are common because these methods require a combination of patch cords with different polarity types.

So what would a proper MPO cable test look like? The answer is simple: Test all 12 fibers – the whole cable – simultaneously and comprehensively (including loss, polarity, etc.). That sort of test capability changes the fiber landscape, enabling installers and technicians to efficiently validate and troubleshoot fiber – flying through the process by tackling an entire 12-fiber cable trunk with the push of a button.

The tools to perform this type of test are just emerging on the market, and promise to reduce the time and labor costs up to 95% over individual fiber tests (according to internal research based on the average list of standard competitive products). Characteristics to look for in such a tool include:
An onboard MPO connector to eliminate the complexity and manual calculations associated with a fan-out cord.
A single “Scan All” test function that delivers visual verification via an intuitive user interface for all 12 MPO fibers in a connector.
Built-in polarity verification for end-to-end connectivity of MPO trunk cables.
“Select Individual Fiber” function that enables the user to troubleshoot a single fiber with more precision.
Demand for fast and reliable delivery of critical applications is driving data center technology to evolve at an ever-increasing pace. And that insatiable need for bandwidth ensures that the integrity of the data center has become inextricably linked to the strength of the fiber cabling infrastructure. The growing use of MPO fiber trunks – and the migration from 10-Gbps to 40/100-Gbps connections – means that it’s time to stop the cumbersome verification of individual fibers. After all, it’s a single MPO connection. You should be able to test it as one.
You can buy fiber optic jumpers with mpo/mpo connectors  from Optoroute now!

What is ROADM?

 

Transmode's ROADM units are a powerful part of the TM-Series enabling cost optimized Flexible Optical Networks based on DWDM technology. Built for dynamic network applications, the ROADM units allow hitless wavelength allocation and future proof traffic design.

Key benefits with Transmode's ROADM units are:

• Multiple use - can be used in bus and ring topologies as well as three- to eight-way mesh nodes with local add-drop
• Available as 2, 4 and 8 individual drop ports providing cost efficient alternatives in both access and metro networks
• Supports cutting edge ROADM node configurations such as colorless, directionless and contentionless nodes, allowing flexible and cost optimized handling of traffic
• Provides add/drop of any wavelength from/to any port, giving maximum flexibility in wavelength allocation
• Built-in Variable Optical Attenuator (VOA) for easier channel power balancing
• Compact design, giving small footprint
• Fully integrated with TM-Series and Transmode Network Manager (TNM)
• Low Power Design and small footprint ensures low total cost of ownership

The 1x2, 1x4 and 1x8 ROADM units are compact solutions for all network topologies. They are fully supported as units in the TM-3000 chassis and work as building blocks for reconfigurable add-drop nodes in up to 8 dimensions.

For more informtion, please visit http://www.optoroute.com.cn/

DWDM Play a significant Role in long-haul optical networks

Dense Wavelength Division Multiplexing (DWDM) is the technology of choice for transporting extremely large amounts of data traffic over metro or long distances in telecom networks. Optical networking and especially the use of DWDM technology has proven to be the optimal way of combining cost efficient transport with advanced functionality, which can cope with the bandwidth explosion from the access network.

DWDM typically has the capability to transport up to 80 channels (wavelengths) in what is known as the Conventional band or C band spectrum, with all 80 channels in the 1550 nm region. DWDM takes advantage of the operating window of the Erbium Doped Fibre Amplifier (EDFA) to amplify the optical channels and extend the operating range of the system to over 1500 kilometers. This denser channel spacing requires tighter control of the wavelengths and therefore cooled lasers as opposed to Coarse Wavelength Division Multiplexing (CWDM) which has broader channel spacing un-cooled lasers.
Transmode uses both DWDM and CWDM technologies as a way of transporting different types of services, e.g. Ethernet, SDH/SONET, and Fibre Channel (FC) for solutions in metro access to regional backhaul networks. DWDM is the variant of WDM technology that provides more functionality in terms of supported distance (for Transmode up to 1500+ km), number of wavelengths and dynamic behavior using ROADM technology.
Both Transmode's TM-Series and TS-Series areDWDM and CWDM agnostic. This means a CWDM network can initially be deployed with either product series, and when required, the network can be simply upgraded to a hybrid CWDM/DWDM network using common cards and pluggable optics. Therefore, by deploying Transmode'sDWDMor CWDM based solutions the lowest possible day 1 cost are enabled without sacrificing the scalability of the network.

For more information on DWDM, please visit http://www.optoroute.com.cn/