DÜSSELDORF -- ECOC 2016 -- Merry Pranksters and Top Gun cadets are dilettantes compared to network operators when it comes to a need for speed, and nowhere do operators need speed more than in data centers and between data centers. The opto-electronics industry is meeting at ECOC in Düsseldorf, Germany, this week to unveil and examine new products and technologies to try to sate the appetite for networking at 100 Gbit/s, 200 Gbit/s, 500 Gbit/s and even "Furthur."
So what's on show? Here's a look at some of the new devices and technologies being unveiled this week, and some of the achievements indicating that accelerating networks to hundreds of gigabits a second is a realistic goal in the foreseeable future.
With internal traffic within data centers (so-called east-west traffic) growing at eye-popping rates, there are any number of bottlenecks, including physical space on server faceplates for connectors. Accelerating the amount of traffic through any given port relieves the problem.
The three have combined AppliedMicro's 16nm FinFET 100Gbit/s PAM4 DSP, BrPhotonics' TFPS modulator, and Macoms's transimpedance amplifier (TIA), and have successfully demonstrated the transmission of 100Gbit/s over a single wavelength.
The companies claim they are the first to enable 400Gbit/s connectivity in a QSFP family of transceivers. They expect commercial deployment will begin in 2017.
There, the standard upgrade path is to switch to multimode ribbon fiber and use 100G SR4 over four parallel data streams, according to Lumentum and Finisar.
But it's possible to go to 100G in several other ways. It's not clear yet if there is a best way, or if different approaches will be useful depending on what a data center already has installed.
The team of Mellanox Technologies Ltd. (Nasdaq: MLNX) and Oclaro Inc. (Nasdaq: OCLR) have combined the latter's 1310nm LR (long reach) SFP28 transceiver line with the former's 1550nm PSM4 transceivers.
The point of connecting 100Gbit/s PSM4 silicon photonics to 25Gbit/s LR transceivers for servers and storage is to enable data center operators to use single mode fiber to interface with 100Gbit/s switches and routers. To put a finer point on it, it provides data center operators a path to scale 25Gbit/s Ethernet networks for east-west data center traffic with links of 100 meters to 2 kilometers.
Mellanox and Oclaro contend that having a single laser and no multiplexing ends up being more cost efficient than approaches that rely on multiplexing multiple wavelengths produced by multiple lasers.
The two companies are demonstrating the compatibility of their 100G SWDM4 optical transceivers. Both have QSFP28 modules based on a common set of specifications. Their demo uses test systems provided by Anritsu Corp. and Viavi Solutions Inc. . All four companies are among the members of the SWDM Alliance.
Using SWDM, multiple data streams from lasers emitting different wavelengths in the 850 nm band are multiplexed into a single fiber at the transmitter and demuxed at the receiver. The two companies explain that this provides a direct, low-cost solution for 100G data rate upgrades for data center customers who have an installed base of duplex multimode fiber (MMF) running at 10G.
ColorChip says its transceiver is interoperable with 100G CWDM4 and CLR4 2Km transceivers as it is identical in the hardware design, and likewise offers a cost effective solution in a small QSFP28 form factor.
The company's products are based on a proprietary waveguide-in-glass PLC-based (planar lightwave circuit) optical platform, coupled with fully automated photonic integration of active and passive optical elements. The company claims the approach has a number of advantages that ultimately result in a reliable, low-cost solution.
Effect Photonics is taking the system on a chip (SoC) approach to optical networking. The company says it combines all optical functions into a single chip, including multiple DBR lasers, low-voltage drive MZ modulators, AWG, and wavelength monitoring.
Effect launched its 100G DWDM transceiver product family half a year ago; it has now moved on. The company, in collaboration with the Technical University of Eindhoven, said it is using PAM-4 to send 500G DWDM signals up to 75 km.
The combination of the optical SoC approach, proprietary packaging, and PAM-4 reduces the cost, power and the physical size of the module. The approach sidesteps the need for long-haul coherent optical transceivers, Effect claims.
Using dual polarization QPSK modulation rather than QAM, however, the HB Micro-ICR extends the reach to 2,000km at 200 Gbit/s, which NeoPhotonics said is double the data rate of standard coherent transmission.
The approach allows network system designers to increase bandwidth by up to a factor of six and dynamically change the reach and modulation format while keeping the number of components unchanged, which the company said helps play into the SDN trend, while greatly reducing the cost per bit.
HiLight Semiconductor said it just received a grant of about $2 million from the European Commission to pursue the development of ultra-low power CMOS ICs for 100G transmission.
HiLight contends that using standard CMOS can save a significant amount of power, especially in data centers.
The company said it expects to have samples available in early 2017.
HiLight recently launched its low power chipset for 10G CMOS SFP+ applications.
Separately, Nokia Corp. (NYSE: NOK) Bell Labs, Deutsche Telekom AG (NYSE: DT) T-Labs, and the Technical University of Munich said they have achieved a 1Tbit/s transmission rate in a field test, by using a new modulation technique that builds on QAM. (See Eurobites: Nokia, DT Hit 1 Tbit/s in Fiber Trial.)
The demonstration shows that the flexibility and performance of optical networks can be maximized when adjustable transmission rates are dynamically adapted to channel conditions and traffic demands, the companies said. (See Nokia, DT Break Terabit Barrier.)
The new modulation technique, called probabilistic constellation shaping (PCS), modifies the probability with which QAM constellation points are used. Traditionally, all constellation points are used with the same frequency, the companies explained. PCS instead uses constellation points with high amplitude less frequently than those with lesser amplitude to transmit signals that, on average, are more resilient to noise and other impairments. This allows the transmission rate to be tailored to ideally fit the transmission channel, delivering up to 30% greater reach.
— Brian Santo, Senior Editor, Components, T&M, Light Reading