COBO acts to bring optics closer to the chip
The goal of COBO, announced at the OFC 2015 show and backed by such companies as Microsoft, Cisco Systems, Finisar and Intel, is to develop a technology roadmap and common specifications for on-board optics to ensure interoperability.
“The Microsoft initiative is looking at the next wave of innovation as it relates to bringing optics closer to the CPU,” says Saeid Aramideh, co-founder and chief marketing and sales officer for start-up Ranovus, one of the founding members of COBO. “There are tremendous benefits for such an architecture in terms of reducing power dissipation and increasing the front panel density.”
On-board optics refers to optical engines or modules placed on the printed circuit board, close to a chip. The technology is not new; Avago Technologies and Finisar have been selling such products for years. But these products are custom and not interoperable.
Placing the on-board optics nearer the chip - an Ethernet switch, network processor or a microprocessor for example - shortens the length of the board’s copper traces linking the two. The fibre from the on-board optics bridges the remaining distance to the equipment’s face plate connector. Moving the optics onto the board reduces the overall power consumption, especially as 25 Gigabit-per-second electrical lanes start to be used. The fibre connector also uses far less face plate area compared to pluggable modules, whether the CFP2, CFP4, QSFP28 or even an SFP+.
“The [COBO] initiative is going to be around defining the electrical interface, the mechanical interface, the power budget, the heat-sinking constraints and the like,” says Aramideh.
To understand why such on-board optics will be needed, Aramideh cites Broadcom’s StrataXGS Tomahawk switch chips used for top-of-rack and aggregation switches. The Tomahawk is Broadcom’s first switch family that use 25 Gbps serialiser/ deserialiser (serdes) and has an aggregate switch bandwidth of up to 3.2 terabit. And Broadcom is not alone. Cavium through its Xplaint acquisition has the CNX880xx line of Ethernet switch chips that also uses 25 Gbps lanes and has a switch capacity up to 3.2 terabit.
“You have 1.6 terabit going to the front panel and 1.6 terabit going to the back panel; that is a lot of traces,” says Aramideh. “If you make this into opex [operation expense], and put the optics close to the switch ASIC, the overall power consumption is reduced and you have connectivity to the front and the back.”
This is the focus of Ranovus, with the OpenOptics MSA initiative. “Scaling into terabit connectivity over short distances and long distances,” he says.
OpenOptics MSA
At OFC, members of the OpenOptics MSA, of which Ranovus and Mellanox are founders, published its WDM specification for an interoperable 100 Gbps WDM standard that will have a two kilometer reach.
The 100 Gigabit standard uses 4x25 Gbps wavelengths but Aramideh says the standard scales to 8, 16 and 32 lanes. In turn, there will also be a 50 Gbps lane version that will provide a total connectivity of 1.6 terabit (32x50 Gbps).
Ranovus has not detailed what modulation scheme it will use to achieve 50 Gbps lanes, but Aramideh says that PAM-4 is one of the options and an attractive one at that. “There are also a lot of chipsets [supporting PAM-4] becoming available,” he says.
Ranovus’s first products will be an OpenOptics MSA optical engine and an QSFP28 optical module. “We are not making any product announcements yet but there will be products available this year,” says Aramideh.
Meanwhile, Ciena has become the sixth member to join the OpenOptics MSA.
OFC 2014 product round-up - Final part
The industry is moving at a clip to fill the void in 100 Gig IEEE standards for 100m to 2km links. Until now, the IEEE 10km 100GBASE-LR4 and the 10x10 MSA have been the interfaces used to address such spans.
But responding to data centre operators, optical players are busy developing less costly, mid-reach MSAs, as was evident at the OFC exhibition and conference, held in San Francisco in March.
Meanwhile, existing IEEE 100 Gigabit standards are skipping to the most compact CFP4 and QSFP28 form factors. The -LR4 standard was first announced in a CFP in 2010, and moved to the CFP2, half the size of the CFP, in 2013. Now, several companies have detailed CFP4 -LR4 products, while Source Photonics has gone one better, announcing the standard in a QSFP28.
The CFP4 is half the size of the CFP2, while the QSFP28 is marginally smaller than the CFP4 but has a lower power consumption: 3.5W compared to the CFP4's 6W.
Timeline of some pluggable announcements at recent OFCs. Source: Gazettabyte
The mid-reach landscape
Several interfaces for mid-reach interconnect were detailed at OFC. And since the show, two MSAs have been detailed: the CWDM4 and the CLR4 Alliance.
At OFC, the OpenOptics MSA backed by Mellanox Technologies and Ranovus, was announced. Skorpios Technologies demoed its CLR4 module that has since become the CLR4 Alliance. And vendors discussed the Parallel Single Mode (PSM4) initiative that was first detailed in January.
Switch vendor Mellanox Technologies and module start-up Ranovus announced the OpenOptics MSA at OFC. The QSFP-based MSA uses a single-mode fibre and WDM transmission around 1550nm to address data centre links up to 2km.
Saeid Aramideh of Ranovus says that the MSA using its laser and silicon photonics technologies will deliver significant cost, power and size advantages {add link}. "But the 1550nm WDM connection is open to any technology," says Aramideh, chief marketing and sales officer at Ranovus. "It does not have to be silicon photonics."
The first MSA product, a 100 Gig QSFP28, uses 4x25 Gig channels. "The channel spacing for the MSA is flexible to be 50GHz or more," says Aramideh. The MSA is scalable to 400 Gig and greater rates. The 100 Gig QSFP28 technology is several months away from sampling.
Skorpios Technologies demonstrated its QSFP28-CLR4 transceiver although the details of the MSA have yet to be detailed. Skorpios is a silicon photonics player and uses heterogenous integration where the lasers, modulators, detectors and optical multiplexer and de-multiplexer are monolithically integrated on one chip.
The PSM4 MSA is another initiative designed to tackle the gap between IEEE short and long reach standards. Backed by players such as Avago Technologies, Brocade, JDSU, Luxtera, Oclaro, and Panduit, the 100 Gig standard is defined to operate in the 1295-1325nm spectral window and will have a reach of at least 500m.
ColorChip demonstrated a 100 Gig (4x25 Gig) QSFP28 with a 2km reach at the show. The design uses uncooled directly modulated lasers to achieve the 3.5W power consumption. Since the show Colorchip is one of the member companies backing the CLR4 Alliance and the demoed QSFP matches the first details of the new MSA's spec.
100GBASE-LR4 moves to CFP4 and QSFP28
The IEEE 100GBASE-LR4 standard is transitioning to the smallest modules. At OFC, vendors detailed the first CFP4s while Source Photonics announced the -LR4 in a QSFP28.
Source Photonics says its transceiver consumes 3.5W. The QSFP28 form factor achieves up to a fourfold increase in face plate density compared to the CFP2: up to 48 modules compared to a dozen CFP2 modules, says the company, which expects first QSFP28 -LR4 samples in mid-2014.
Meanwhile, Avago Technologies, Finisar, Fujitsu Optical Components and JDSU all detailed their first CFP4 -LR4 modules.
JDSU says that when it developed the optics for its CFP2 -LR4, it was already eyeing the transition to the CFP4 and QSFP28 form factors. To achieve the -LR4 spec in the 6W CFP4, a key focus are the clock data recover (CDR), driver and trans-impedance amplifier chips. "A decent amount of the power consumption is wrapped up in the ICs that do the CDR and a variety of the digital functions behind the photonics," says Brandon Collings, JDSU's CTO for communications and commercial optical products. JDSU expects general availability of its CFP4 -LR4 later this year.
Finisar's -LR4 is its second CFP4 product; at ECOC 2013 it showcased a 100m, 100GBASE-SR4 CFP4. Finisar says its -LR4 uses distributed feedback (DFB) lasers and consumes 4.5W, well within the CFP4's 6W power profile. At OFC, the CFP4 was demonstrated working with CFP2 and CFP -LR4 modules. Finisar's CFP4 will sample later this year.
Avago announced availability of its -LR4 transmit optical subassembly (TOSA) and receive optical subassembly (ROSA) products for the CFP4, along with its CFP4 module which it says will be available next year. Fujitsu Optical Components also used OFC to demo its CFP4 -LR4.
40km Extended Reach Lite
Oclaro and Finisar detailed a tweak to the 100 Gig Extended Reach standard: the 40km, 100GBASE-ER4.
The IEEE standard uses a power-hungry semiconductor optical amplifier (SOA) prior to the PIN photodetector to achieve 40km. The module vendors have proposed replacing the SOA and PIN with an avalanche photo diode (APD) and external forward error correction to reduce the power consumption while maintaining the optical link budget. The changed spec is dubbed 100GBASE-ER4 Lite.
"Trying to achieve the power envelopes required for the CFP4 and QSFP28 using SOAs is going to be too hard," says Kevin Granucci, vice president of strategy and marketing at Oclaro.
Oclaro demonstrated a ER4-Lite in a CFP2. The module supports 100 Gigabit Ethernet and the Optical Transport Network (OTN) OTU-4 rates, and consumes less than 9W. "We are using the CFP2 as the first proof-of-concept," says Granucci. "For the 6W CFP4 and the 3.5W QSFP28, we think this is the only solution available."
At OFC Finisar demonstrated the link's feasibility, which it refers to as ER4f, using four 28 Gig lasers and four 28 Gig APDs.
Oclaro says it is seeing customer interest in the ER4 Lite, and points out that there are many 10 Gig 40km links deployed, especially in China. "The ER4 Lite will provide an update path to 100 Gig," says Granucci.
VCSELs: serial 40 Gig and the 400 Gig CDFP
Finisar showcased a VCSEL operating at 40 Gig at OFC. State-of-the-art VCSEL interfaces run up to 28 Gig. Finisar's VCSEL demonstration was to show the commercial viability of higher-speed VCSELs for single channel or parallel-array applications. "We believe that VCSELs have not run out of steam," says Rafik Ward, vice president of marketing at Finisar. The 40 Gig VCSEL demonstration used non-return-to-zero (NRZ) signalling, "no higher-order modulation is being used", says Ward.
IBM T.J.Watson Research Center has published an IEEE paper with Finisar involving a 56Gbps optical link based on an 850nm VCSEL.
Finisar also demonstrated an CDFP-based active optical cable. The CDFP is a 400 Gig MSA that uses 16 x 25 Gig VCSEL channels in each direction. Such an interface will address routing, high-performance computing and proprietary interface requirements, says Finisar. The demonstration showcased the technology; Finisar has yet to announce interface products or reaches.
Short reach 100G and 4x16 Gig Fibre QSFPs
Avago Technologies announced a 100GBASE-SR4 implemented using the QSFP28. Avago's I Hsing Tan, segment marketing manager for Ethernet and storage optical transceivers, says there has been a significant ramp in data centre demand for the 40GBASE-SR4 QSFP+ in the last year. "Moving to the next generation, the data centre operator would like to keep the same [switch] density as the QSFP+, and the QSFP28 MSA offers the same form factor," he says.
The QSFP28 differs from the QSFP+ is that its electrical connector is upgraded to handle 28 Gigabit-per-lane data rates. Avago says the -SR4 module will be generally available next year.
Avago also announced a 4x16 Gigabit Fibre Channel QSFP+ transceiver. The industry is transitioning from 8 to 16 Gig Fibre Channel, says Avago, and this will be followed by 32 Gig serial and 4x32 Gig Fibre Channel modules.
The company has announced a 4x16 Gig QSFP+ to continue the increase in platform channel density while the industry transitions from 16 to 32 Gig Fibre Channel. "This solution is going to provide the switch vendor a 3x increase in density at half the power dissipation per channel for 16 Gig Fibre Channel, before the 32 bit Fibre Channel come to maturity in three to five years," says Tan.
Avago has just announced that it has shipped over half a million QSFP+ modules.
Optical engines
TE Connectivity announced its 25 Gig-per-channel optical engine technology. The Coolbit optical engine will be included in four TE Connectivity products planned for this year: 100 Gig QSFP28 active optical cables (AOCs), 100 Gig QSFP28 transceivers, 300 Gig mid-board optical modules, and 400 Gig CDFP AOCs.
Meanwhile, Avago's MiniPod and MicroPod optical engine products now have a reach of 550m when coupled with Corning's ClearCurve OM4 fibre.
"This allows customers in the data centre to go a little bit further and not have to go to single-mode fibre," says Sharon Hall, product line manager for embedded optics at Avago.
For Part 1, click here
Further reading:-
TE Connectivity White Paper: End-to-end Communications with Fiber Optic Technologies, click here
LightCounting: Reflections on OFC 2014: The industry is approaching a critical junction, click here
Ovum at OFC 2014, click here
LightWave OFC 2014 Podcast, click here
Ethernet Alliance Blog: OFC 2014 show and best in class, click here
OFC 2014 industry reflections - Part 2
The high cost of 100 Gigabit Ethernet client modules has been a major disappointment to me as it has slowed adoption
Joe Berthold, Ciena
Joe Berthold, vice president of network architecture at Ciena.
OFC 2014 was another great event, with interesting programmes, demonstrations and papers presented. A few topics that really grabbed my interest were discussions around silicon photonics, software-defined networking (SDN) and 400 Gigabit Ethernet (GbE).
The intense interest we saw at last year’s OFC around silicon photonics grew this year with lots of good papers and standing-room-only sessions. I look forward to future product announcements that deliver on the potential of this technology to significantly reduce cost of interconnecting systems over modest distances. The high cost of 100GbE client modules has been a major disappointment to me as it has slowed adoption.
Another area of interest at this year’s show was the great deal of experimental work around SDN, some more practical than others.
I particularly liked the reviews of the latest work under the DARPA-sponsored CORONET programme, whose Phase 3 focused on SDN control of multi-layer, multi-vendor, multi-data centre cloud networking across wide area networks.
In particular, there were talks from three companies I noted: Anne Von Lehman of Applied Communication Sciences, the prime contractor, provided a good program overview; Bob Doverspike of AT&T described a very extensive testbed using equipment of the type currently deployed in AT&T’s network, as well as two different processing and storage virtualisation platforms; and Doug Freimuth of IBM described its contributions to CORONET including an OpenStack virtualisation environment, as well as other IBM distributed cloud networking research.
All the action on rates above 100 Gig lies with the selection of client signals. 400 Gig seems to have the major mindshare but there are still calls for flexible rate clients and Terabit clients.
One thing I enjoyed about these talks was that they described an approach to SDN for distributed data centre networking that is pragmatic and could be realised soon.
I also really liked a workshop held on the Sunday on the question whether SDN will kill GMPLS. While there was broad consensus that GMPLS has failed in delivering on its original turn-of-the-century vision of IP routers control of multi-layer, multi-domain networks, most speakers recognised the value distributed control planes have in simplifying and speeding the control of single layer, single domain networks.
What I took away was that single layer distributed control planes are here to stay as important network control functions, but instead will work under the direction of an SDN network controller.
As we all know, 400 Gigabit dense wavelength division multiplexing (DWDM) is here from the technology perspective, but awaiting standardisation of the 400 Gig Ethernet signal from the IEEE, and follow-on work by the ITU-T on signal mapping to OTN. In fact, from the perspective of DWDM transmission systems, 1 Terabit-per-second systems can be had for the asking.
All the action on rates above 100 Gig lies with the selection of client signals. 400 Gig seems to have the major mindshare but there are still calls for flexible rate clients and Terabit clients.
One area that received a lot of attention, with many differing points of view, was the question of the 400GbE client. As the 400GbE project begins soon in the IEEE, it is time to take a lesson from the history of the 100 Gig client modules and do better.
Let us all agree that we don’t need 400 Gig clients until they can do better in cost, face plate density, and power dissipation than the best 100 Gig modules that will exist then.
The first 100 Gig DWDM transceivers were introduced in 2009. It is now 2014 and 100 Gig is the transmission rate of choice for virtually all high capacity DWDM network applications, with a strong economic value proposition versus 10 Gig. Yet the industry has not yet managed to achieve cost/bit parity between 100 Gig and 10 Gig clients - far from it!
Last year's OFC, we saw many show floor demonstrations of CFP2 modules. They promise lower costs, but evidence of their presence in shipping products is still lacking. At the exhibit this year we saw 100 Gig QSFP28 modules. While progress is slow, the cost of the 100 Gig client module continues to result in many operators favouring 10 Gig handoffs to their 100 Gig optical networking systems.
Let us all agree that we don’t need 400 Gig clients until they can do better in cost, face plate density, and power dissipation than the best 100 Gig modules that will exist then. At this juncture the 100 Gig benchmark we should be comparing 400 Gig to is a QSFP28 package.
Lastly, last year we heard about the launch of an OIF project to create a pluggable analogue coherent optical module. There were several talks that referenced this project, and discussed its implications for shrinking size and supporting higher transceiver card density.
Broad adoption of this component will help drive down costs of coherent transceivers, so I look forward to its hearing about its progress at OFC 2015.
Daryl Inniss, vice president and practice leader, Ovum.
There was no shortage of client-side announcements at OFC and I’ve spent time since the conference trying to organise them and understand what it all means.
I’m tempted to say that the market is once again developing too many options and not quickly agreeing on a common solution. But I’m reminded that this market works collaboratively and the client-side uncertainty we’re seeing today is a reflection of a lack of market clarity.
Let me describe three forces affecting suppliers:
The IEEE 100GBASE-xxx standards represent the best collective information that suppliers have. Not surprisingly, most vendors brought solutions to OFC supporting these standards. Vendors sharpened their products and focused on delivering solutions with smaller form factors and lower power consumption. Advances in optical components (lasers, TOSAs and ROSAs), integrated circuits (CDRs, TIAs, drivers), transceivers, active optical cables, and optical engines were all presented. A promising and robust supply base is emerging that should serve the market well.
A second driver is that hyperscale service providers want a cost-effective solution today that supports 500m to 2km. This is non-standard and suppliers have not agreed on the best approach. This is where the market becomes fragmented. The same vendors supporting the IEEE standard are also pushing non-standard solutions. There are at least four different approaches to support the hyperscale request:
- Parallel single mode (PSM4) where an MSA was established in January 2014
- Coarse wavelength division multiplexing—using uncooled directly modulated lasers and single mode fibre
- Dense wavelength division multiplexing—this one just emerged on the scene at OFC with Ranovus and Mellanox introducing the OpenOptics MSA
- Complex modulation—PAM-8 for example and carrier multi-tone.
Admittedly, the presence of this demand disrupts the traditional process. But I believe the suppliers’ behavior reflects their unhappiness with the standardisation solution.
The good news is these approaches are using established form factors like the QSFP. And silicon photonic products are starting to emerge. Suppliers will continue to innovate.
Ambiguity will persist but we believe that clarity will ultimately prevail.
The third issue lurking in the background is knowledge that 400 Gig and one Terabit will soon be needed. The best-case scenario is to use 100 Gig as a platform to support the next generation. Some argue for complex modulation as you reduce the number of optical components thereby lowering cost. That’s good but part of the price is higher power consumption, an issue that is to be determined.
Part of today’s uncertainty is whether the standard solution is suitable to support the market to the next generation. Sixteen channels at 25 Gig is doable but feels more like a stopgap measure than a long-term solution.
These forces leave suppliers innovating in search of the best path forward. The approaches and solutions differ for each vendor. Timing is an issue too with hyperscale looking for solutions today while the mass market may be years away.
We believe that servers with 25 Gig and/ or 40 Gig ports will be one of the catalysts to drive the mass market and this will not start until about 2016. Meanwhile, each vendor and the market will battle for the apparent best solution to meet the varying demands. Ambiguity will persist but we believe that clarity will ultimately prevail.