Industry in a flurry of mid-reach MSA announcements
Another day, another multi-source agreement.
The CLR4 Alliance is the latest 100 Gig multi-source agreement (MSA) to address up-to-2km links in the data centre. The 100 Gig CLR4 Alliance is backed by around 20 companies including data centre operators, equipment vendors, optical module and component players and chip makers.
The table provides a summary of the latest MSAs and how they relate to the IEEE 100 Gigabit client interface standards. Source: Gazettabyte.
The announcement follows in the footsteps of the CWDM4 MSA, announced at the start of the week. The CWDM4 is another 100 Gig single-mode fibre MSA backed by optical module makers, Avago Technologies, Finisar, JDSU and Oclaro.
The two MSAs are the latest of several announced interfaces - three in the last three weeks - to tackle mid-reach distances from 100m-plus to 2km. The MSAs reflect an industry need to fill the void in the IEEE standards: the -SR4 multimode standard, with its 100m reach, and the 10km -LR4 that is seen as over-specified for data centre requirements.
Below is a discussion of the recent data centre MSAs
The PSM4 MSA
The PSM4 MSA is a four-channel parallel single mode interface that uses eight- or 12-fibre cabling based on the MTP/MPO optical connectors. The PSM4 uses simpler optics than the 10km IEEE 100GBASE-LR4 and the shorter-reach 2km offshoot, the CWDM4 MSA, and thus promises lower cost. But this is at the expense of using eight fibres and more expensive connectors compared to the single-mode CWDM4.
The PSM4 is expected to have a reach of at least 500m; above that the cost of the fibre becomes the dominant factor. "Note that a 500m PMD [physical medium dependent layer] at 100 Gig was an objective of the IEEE 802.3bm group but it did not happen, so the industry is defining products that fill the gap," says Dale Murray, principal analyst at LightCounting Market Research.
The PSM4 MSA was first detailed in January and includes such members as Avago Technologies, Brocade, JDSU, Luxtera, Oclaro and Panduit.
The 100 Gig CLR4 Alliance
The 100 Gig CLR4 MSA is backed by companies including ebay; equipment vendors Arista Networks, Brocade, Dell, Fujitsu, HP, and Oracle; silicon photonics players Aurrion, Intel, Skorpios Technologies (Oracle is also a proponent of silicon photonics); optical module and component players ColorChip, Kaiam, Oclaro, Oplink, NeoPhotonics; and chip vendors, Netronome and Semtech.
The 100 Gig standard is based on a QSFP form factor module and uses two single mode fibres - 1 send and 1 receive - for duplex communications. The MSA has a 2km reach and uses coarse wavelength-division multiplexing (CWDM). The 8.5mm x18mm x 72mm QSFP has a maximum power consumption of 3.5W and enables a port density of 36 modules on a face plate of a 1 rack unit card, or 3.6 Terabits overall.
At the recent OFC show, Skorpios Technologies demonstrated a QSFP28-CLR4. The silicon photonics player said its module was based on a single-chip that integrates the lasers, modulators, detectors and optical multiplexer and de-multiplexer, to deliver significant size, cost and power benefits. It also said the transceiver achieved a reach of 10km, putting its CLR4 on a par with the IEEE -LR4
At OFC ColorChip announced its iLR4 which is a QSFP28 with a 2km although, like with Skorpios, this was before the 100G CLR4 Alliance launch.
The CWDM4 MSA
The CWDM4 MSA also uses 4-channel CWDM optics and two-fibre cabling. The CWDM4 is being promoted as a complement to the PSM4. "It is MSA-based and has a 2km target," says Murray. "This is an LR4 with relaxed specs; it has no thermal electric cooler but uses the same wavelengths."
"From the link solution point of view, the PSM4 may be more cost effective than the CWDM4 up to 200m-300m," says I-Hsing Tan, segment marketing manager for Ethernet and storage optical transceivers at Avago. "But CWDM4 is for sure the winner beyond 200m and can be more cost effective than the 100GBASE-LR4 solution up to 2km."
Companies backing the CWDM4 MSA include Avago Technologies, Finisar, JDSU and Oclaro.
The OpenOptics MSA
The OpenOptics MSA was launched at OFC by Mellanox Technologies and Ranovus. The MSA uses 1550nm optics and DWDM. The first implementation will be a 100G QSFP28 module and the distance it will address is up to 2km. The MSA will also support future 400 Gig and greater interface speeds.
The degree of acceptance of the OpenOptics MSA is still to be determined compared to the more broadly backed CWDM4.
Other developments
The CLR4 Alliance may not be the final word regarding MSA announcements for the data centre.
Work is ongoing to use advanced modulation for data centre links, such as PAM-8 and carrier multi-tone.
Both Ciena's Joe Berthold and Ovum's Daryl Inniss address the importance of client-side interfaces and whether the rush to announce new MSAs is beneficial overall.
The story was first published on April 1st and has been updated to include the CLR4 Alliance MSA.
OFC/NFOEC 2013 product round-up - Part 1
Part 1: Client-side transceivers
- First CFP2 single-mode and multi-mode transceiver announcements
- Cisco Systems unveils its CPAK module
- 100 Gigabit QSFPs from Kotura and Luxtera
- CFP2 and 40km CFP 10x10 MSA modules
- Infiniband FDR and 'LR4 superset' QSFPs
The recent OFC/NFOEC exhibition and conference held in Anaheim, California, saw a slew of optical transceiver announcements. The first CFP2 client-side products for single-mode and multi-mode fibre were unveiled by several companies, as was Cisco Systems' in-house CPAK transceiver.
The CFP2 is the pluggable form factor that follows the first generation CFP. The CFP MSA announced the completion of the CFP2 specification at the show, while several vendors including Avago Technologies, Finisar, Fujitsu Optical Components, NeoPhotonics, Oclaro and Oplink Communications detailed their first CFP2 products.
The 40 and 100 Gigabit CFP2 is half the size of the CFP, enabling at least a doubling of the CFP2 transceivers on a faceplate compared to four CFPs (see table below). The CFP2 is also future-proofed to support 200 and 400Gbps (See first comment at bottom of CFP2 story).
Another difference between the CFP and the CFP2 is that the CFP2 uses a 4x25Gbps electrical interface. Accordingly, the CFP2 does not need the 'gearbox' IC that translates between ten, 10 Gigabit-per-second (Gbps) lanes to four, 25Gbps electrical lanes that interface to the 4x25/28Gbps optics. Removing the gearbox IC saves space and reduces the power consumption by several watts.
The industry has long settled on the SFP+ at 10Gbps while the QSFP has become the 40Gbps form factor of choice. With 100Gbps still in its infancy, transceiver vendors are pursuing several client-side interfaces. Much work will be needed to reduce the size, power consumption and cost of 100Gbps interfaces before the industry settles on a single pluggable form factor for the single-mode and multi-mode standards.
CFP2 announcements
Finisar demonstrated two CFP2 modules, one implementing the IEEE 100GBASE-LR4 10km standard and the other, the IEEE 100GBASE-SR10 100m multi-mode standard. The company is using directly-modulated, distributed feedback (DFB) lasers for its CFP2 LR4. In contrast, the CFP module uses more expensive, electro-absorption modulator lasers (EMLs). Finisar demonstrated interoperability between the two LR4 modules, an EML-based CFP and a DFB-based CFP2, at the show.
* An ER4 CFP2 is under development
** Oclaro disclosed indium phosphide components for a future CFP2 line side pluggable
Using directly modulated lasers also reduces the power consumption, says Finisar. Overall, the CFP2 LR4 consumes 7W compared to a 24W first-generation CFP-based LR4.
"We can migrate these [directly modulated laser] designs to a single quad 28 Gig photonic integrated circuit TOSA," says Rafik Ward, Finisar's vice president of marketing. "Likewise on the receive [path], there will be a quad 28 Gig ROSA." The TOSA refers to a transmitter optical sub-assembly while the ROSA is the receiver equivalent. Ward says the CFP2s will be in production this year.
Several module and chip makers took part in the Optical Internetworking Forum's (OIF) multi-vendor demonstration of its 4x25 Gigabit chip-to-module electrical interface, the CEI-28G-VSR. The demonstration included CFP2 LR4s from Finisar and from Oclaro as well as Luxtera's 100Gbps shorter reach module in a QSFP28. Oclaro's CFP2 is expected to be in production in the third quarter of 2013.
Another standard implemented in the CFP2 is the 100GBASE-SR10 multi-mode standard. Avago Technologies and Finisar both detailed CFP2 SR10 modules. The SR10 uses 10 VCSELs, each operating at 10Gbps. The SR10 can be used as a 100Gbps interface or as 10 independent 10Gbps channels.
The CFP2 SR10 can be interfaced to 10 Gigabit Ethernet (GbE) SFP+ modules or combinations of 10GbE SFP+ and 40GbE QSFPs. "What people are looking for using the CFP2 multi-mode module is not only for the 100 Gig Ethernet application but interoperability with 40 Gig Ethernet as well as 10 Gig Ethernet modules," says I Hsing Tan, Ethernet segment marketing manager in the fibre optics product division at Avago.
The SR10 electrical interface specification supports retiming and non-retiming options. The Avago CFP2 module includes clock data recovery ICs that can be used for retiming if needed or bypassed. The result is that Avago's CFP2 SR10 consumes 4-6W, depending on whether the clock data recovery chips are bypassed or used.
Meanwhile, NeoPhotonics became the first company to announce the 10x10 MSA in a CFP2.
NeoPhotonics has not detailed the power consumption but says the 10x10Gbps CFP2 is lower than the CFP since all of the chips - photonic and electrical - are a newer generation and much work has gone into reducing the power consumption.
"Demand is quite strong for the 10x10 solution," says Ferris Lipscomb, vice president of marketing at NeoPhotonics. "The CFP2 version is being developed, and we expect strong demand there as well."
The key advantage of the 10x10-based solution over a 4x25Gbps design is cost, according to NeoPhotonics. "10x10 enjoys the volume and maturity of 10 Gig, and thus the cost advantage," says Lipscomb. "We believe the 10x10 CFP2 will follow the trend of the 10x10 MSA CFP and will offer a significant cost advantage over CFP2 LR4-based solutions."
Cisco's CPAK
Cisco finally showed its in-house silicon photonics-based CPAK transceiver at OFC/NFOEC. The CPAK is the first product to be announced following Cisco's acquisition of silicon photonics player, LightWire.
Cisco says the CPAK is more compact than the CFP2 transceiver with the company claiming that 12 or more transceivers will fit on a faceplate. "While the industry is leapfrogging the CFP with the CFP2, our CPAK leapfrogs the CFP2 because it is much more efficient from a size and power consumption perspective," says Sultan Dawood, a marketing manager at Cisco.
Vendors backing the CFP2 stress that the CPAK is only slighter smaller than the MSA module. "The CFP2 and the CPAK are both interim form factors pending when the CFP4 becomes available." says Avago's Tan. "Any product [like the CFP2] governed by an MSA is going to see strong market adoption."
Cisco's CPAK transceiver Source: Cisco
The CFP4 specification is still being worked on but 16 CFP4s will fit on a faceplate and the transceiver is scheduled for the second half of 2014.
At OFC, Cisco demonstrated the CPAK implementing the 100GBASE-LR4 and -SR10 standards. The CPAK transceiver will be generally available in the summer of 2013, says Cisco.
CFP
Oplink Communication and hybrid integration specialist, Kaiam, showed a 100Gbps 10x10 MSA CFP implementing a 40km extended reach.
The 10x10 40km CFP is for connecting data centres and for broadband backhaul applications. The CFP electro-absorption modulator lasers coupled to a wavelength multiplexer make up the TOSA while the ROSA comprises avalanche photodiode receivers and a demultiplexer. Samples will be available in the second quarter of 2013, with production starting in the third quarter.
Source Photonics announced a second-generation 100GBASE-LR4 CFP with a power consumption of 12-14W.
Meanwhile, Effdon Networks detailed its first 100Gbps product, a CFP with a reach of 80km. Until now 100Gbps CFPs have been limited largely to 10km LR4 while the first 100Gbps CFPs with a reach of 80km or greater being 4x25Gbps direct-detection designs that can include specialist ICs.
100 Gig QSFP
Luxtera and Kotura, both detailed 100 Gigabit QSFPs that use their respective silicon photonics technology. The Kotura design uses two chips, has a reach of 2km and is a four-channel wavelength-division multiplexing (WDM) design while the Luxtera design is a four-channel integrated transceiver that uses a single laser and is tailored for 500m although Luxtera says it can achieve a 2km reach.
40 Gigabit Ethernet and Infiniband FDR
Avago Technologies announced that its eSR4 40 Gigabit Ethernet (GbE) QSFP+ has a reach of up to 550m, beyond the reach specified by the IEEE 40GBASE-SR4 standard. The eSR4 supports 40GbE or four independent 10GbE channels. When used as a multi-channel 10GbE interface, the QSFP+ interfaces to various 10GbE form factors such as X2, XFP and SFP+, It can also interface to a 100GbE CFP2, as mentioned.
Avago first announced the eSR4 QSFP+ with a reach of 300m over OM3 multi-mode fibre and 400m over OM4 fibre. The eSR4 now extends the reach to a guaranteed 550m when used with specific OM4 fibre from fibre makers Corning, Commscope and Panduit.
The extended reach is needed to address larger data centres now being build, as well as support flatter switch architectures that use two rather than three tiers of switches, and that have greater traffic flowing between switches on the same tier.
Avago says data centre managers are moving to deploy OM4 fibre. "The end user is going to move from OM3 to OM4 fibre for future-proofing purposes," says Tan. "The next-generation 32 Gig Fibre Channel and 100 Gigabit Ethernet are focussing on OM4 fibre."
Meanwhile, ColorChip showed its 56Gbps QSFP+ implementing the FDR (Fourteen Data Rate) 4x Infiniband standard as part of a Mellanox MetroX long-haul system demonstration at the show.
Finisar also demonstrated a 40Gbps QSFP using four 1310nm VCSELs. The result is a QSFP with a 10km reach that supports a 40Gbps link or four, 10Gbps links when used in a 'breakout' mode. The existing 40GBASE-LR4 standard supports a 40Gbps link only. Finisar's non-standard implementation adds a point-to-multipoint configuration.
"A single form factor port can be used not only for 40 Gig but also can enable higher density 10 Gig applications than what you can do with SFP+," says Ward.
Kaiam detailed a 40Gbps QSFP+ ER4 transceiver having a 40km reach. The QSFP+ transceiver has the equivalent functionality of four DML-based SFP+s fixed on a coarse WDM grid, and includes a wavelength multiplexer and de-multiplexer.
For OFC/NFOEC 2013 - Part 2, click here
Further reading
LightCounting: OFC/NFOEC review: news from the show floor, click here
Ovum: Cisco hits both show hot buttons with silicon photonics for 100G, click here
Aurrion mixes datacom and telecom lasers on a wafer
"There is an inevitability of the co-mingling of electronics and optics and we are just at the beginning"
Eric Hall, Aurrion
Aurrion's long-term vision for its heterogeneous integration approach to silicon photonics is to tackle all stages of a communication link: the high-bandwidth transmitter, switch and receiver. Heterogeneous integration refers to the introduction of III-V material - used for lasers, modulators and receivers - onto the silicon wafer where it is processed alongside the silicon using masks and lithography.
In a post-deadline paper given at OFC/NFOEC 2013, the fabless start-up detailed the making of various transmitters on a silicon wafer. These include tunable lasers for telecom that cover the C- and L-bands, and uncooled laser arrays for datacom.
The lasers are narrow-linewidth tunable devices for long-haul coherent applications. According to Aurrion, achieving a narrow-linewidth laser typically requires an external cavity whose size makes it difficult to produce a compact design when integrated with the modulator.
Having a tunable laser integrated with the modulator on the same silicon photonics platform will enable compact 100 Gigabit coherent pluggable modules. "The 100 Gig equivalent of the tunable XFP or SFP+," says Eric Hall, vice president of business development at Aurrion.
Hall admits that traditional indium-phosphide laser manufacturers will likely integrate tunable lasers with the modulator to produce compact narrow-linewidth designs. "There will be other approaches but it is exciting that we can now make this laser and modulator on this platform," says Hall. "And it becomes very exciting when you make these on the same wafer as high-volume datacom components."
Aurrion's vision of a coherent transmitter and a 16-laser array made on the same wafer. Source: Aurrion
The wafer's datacom devices include a 4-channel laser array for 100GBASE-LR4 10km reach applications and a 400 Gigabit transmitter design comprising 2x8 wavelength division multiplexing (WDM) arrays for a 16x25Gbps design, each laser spaced 200GHz apart. These could be for 10km or 40km applications depending on the modulator used. "These arrays are for uncooled applications," says Hall. "The idea is these don't have to be coarse WDM but tighter-spaced WDM that hold their wavelength across 20-80oC."
Coarse WDM-based laser arrays do not require a thermo-electric cooler (TEC) but the larger spacing of the wavelengths makes it harder to design beyond 100 Gigabit, says Hall: "Being able to pack in a bunch of wavelengths yet not need a TEC opens up a lot of applications."
Such lasers coupled with different modulators could also benefit 100 Gigabit shorter-reach interfaces currently being discussed in the IEEE, including the possibility of multi-level modulation schemes, says the company.
Aurrion says it is seeing the trend of photonics moving closer to the electronics due to emerging applications.
"Electronics never really noticed photonics because it was so far away and suddenly photonics has encroached into its personal space," says Hall. "There is an inevitability of the co-mingling of electronics and optics and we are just at the beginning."
OFC/NFOEC 2013 to highlight a period of change
Next week's OFC/NFOEC conference and exhibition, to be held in Anaheim, California, provides an opportunity to assess developments in the network and the data centre and get an update on emerging, potentially disruptive technologies.
Source: Gazettabyte
Several networking developments suggest a period of change and opportunity for the industry. Yet the impact on optical component players will be subtle, with players being spared the full effects of any disruption. Meanwhile, industry players must contend with the ongoing challenges of fierce competition and price erosion while also funding much needed innovation.
The last year has seen the rise of software-defined networking (SDN), the operator-backed Network Functions Virtualization (NFV) initiative and growing interest in silicon photonics.
SDN has already being deployed in the data centre. Large data centre adopters are using an open standard implementation of SDN, OpenFlow, to control and tackle changing traffic flow requirements and workloads.
Telcos are also interested in SDN. They view the emerging technology as providing a more fundamental way to optimise their all-IP networks in terms of processing, storage and transport.
Carrier requirements are broader than those of data centre operators; unsurprising given their more complex networks. It is also unclear how open and interoperable SDN will be, given that established vendors are less keen to enable their switches and IP routers to be externally controlled. But the consensus is that the telcos and large content service providers backing SDN are too important to ignore. If traditional switching and routers hamper the initiative with proprietary add-ons, newer players will willing fulfill requirements.
Optical component players must assess how SDN will impact the optical layer and perhaps even components, a topic the OIF is already investigating, while keeping an eye on whether SDN causes market share shifts among switch and router vendors.
The ETSI Network Functions Virtualization (NFV) is an operator-backed initiative that has received far less media attention than SDN. With NFV, telcos want to embrace IT server technology to replace the many specialist hardware boxes that take up valuable space, consume power, add to their already complex operations support systems (OSS) while requiring specialist staff. By moving functions such as firewalls, gateways, and deep packet inspection onto cheap servers scaled using Ethernet switches, operators want lower cost systems running virtualised implementations of these functions.
The two-year NFV initiative could prove disruptive for many specialist vendors albeit ones whose equipment operate at higher layers of the network, removed from the optical layer. But the takeaway for optical component players is how pervasive virtualisation technology is becoming and the continual rise of the data centre.
Silicon photonics is one technology set to impact the data centre. The technology is already being used in active optical cables and optical engines to connect data centre equipment, and soon will appear in optical transceivers such as Cisco Systems' own 100Gbps CPAK module.
Silicon photonics promises to enable designs that disrupt existing equipment. Start-up Compass-EOS has announced a compact IP core router that is already running live operator traffic. The router makes use of a scalable chip coupled to huge-bandwidth optical interfaces based on 168, 8 Gigabit-per-second (Gbps) vertical-cavity surface-emitting lasers (VCSELs) and photodetectors. The Terabit-plus bandwidth enables all the router chips to be connected in a mesh, doing away with the need for the router's midplane and switching fabric.
The integrated silicon-optics design is not strictly silicon photonics - silicon used as a medium for light - but it shows how optics is starting to be used for short distance links to enable disruptive system designs.
Some financial analysts are beating the drum of silicon photonics. But integrated designs using VCSELs, traditional photonic integration and silicon photonics will all co-exist for years to come and even though silicon photonics is expected to make a big impact in the data centre, the Compass-EOS router highlights how disruptive designs can occur in telecoms.
Market status
The optical component industry continues to contend with more immediate challenges after experiencing sharp price declines in 2012.
The good news is that market research companies do not expect a repeat of the harsh price declines anytime soon. They also forecast better market prospects: The Dell'Oro Group expects optical transport to grow through 2017 at a compound annual growth rate (CAGR) of 10 percent, while LightCounting expects the optical transceiver market to grow 50 percent, to US $5.1bn in 2017. Meanwhile Ovum estimates the optical component market will grow by a mid-single-digit percent in 2013 after a contraction in 2012.
In the last year it has become clear how high-speed optical transport will evolve. The equipment makers' latest generation coherent ASICs use advanced modulation techniques, add flexibility by trading transport speed with reach, and use super-channels to support 400 Gigabit and 1 Terabit transmissions. Vendors are also looking longer term to techniques such as spatial-division multiplexing as fibre spectrum usage starts to approach the theoretical limit.
Yet the emphasis on 400 Gigabit and even 1 Terabit is somewhat surprising given how 100 Gigabit deployment is still in its infancy. And if the high-speed optical transmission roadmap is now clear, issues remain.
OFC/NFOEC 2013 will highlight the progress in 100 Gigabit transponder form factors that follow the 5x7-inch MSA, 100 Gigabit pluggable coherent modules, and the uptake of 100 Gigabit direct-detection modules for shorter reach links - tens or hundreds of kilometers - to connect data centres, for example.
There is also an industry consensus regarding wavelength-selective switches (WSSes) - the key building block of ROADMs - with the industry choosing a route-and-select architecture, although that was already the case a year ago.
There will also be announcements at OFC/NFOEC regarding client-side 40 and 100 Gigabit Ethernet developments based on the CFP2 and CFP4 that promise denser interfaces and Terabit capacity blades. Oclaro has already detailed its 100GBASE-LR4 10km CFP2 while Avago Technologies has announced its 100GBASE-SR10 parallel fibre CFP2 with a reach of 150m over OM4 fibre.
The CFP2 and QSFP+ make use of integrated photonic designs. Progress in optical integration, as always, is one topic to watch for at the show.
PON and WDM-PON remain areas of interest. Not so much developments in state-of-the-art transceivers such as for 10 Gigabit EPON and XG-PON1, though clearly of interest, but rather enhancements of existing technologies that benefit the economics of deployment.
The article is based on a news analysis published by the organisers before this year's OFC/NFOEC event.
OneChip Photonics targets the data centre with its PICs
OneChip Photonics is developing integrated optical components for the IEEE 40GBASE-LR4 and 100GBASE-LR4 interface standards.
The company believes its photonic integrated circuits (PICs) will more than halve the cost of the 40 and 100 Gigabit 10km-reach interfaces, enough for LR4 to cost-competitively address shorter reach applications in the data centre.
"I think we can cut the price [of LR4 modules] by half or better”
Andy Weirich, OneChip Photonics
The products mark an expansion of the Canadian startup's offerings. Until now OneChip has concentrated on bringing PIC-based passive optical network (PON) transceivers to market.
LR4 PICs
The startup is developing separate LR4 transmitter and receiver PICs. The 40 and 100GBASE-LR4 receivers are due in the third quarter of 2012, while the transmitters are expected by the year end.
The 40GBASE-LR4 receiver comprises a wavelength demultiplexer - a 4-channel arrayed waveguide grating (AWG) - and four photo-detectors operating around 1300nm. A spot-size converter - an integrated lens - couples the receiver's waveguide's mode field to the connecting fibre.
"[Data centre operators] are saying that they are having to significantly bend out of shape their data centre architecture to accommodate even 300m reaches”
The 40GBASE-LR4 transmitter PIC comprises four directly-modulated distributed feedback (DFB) lasers while the 100GBASE-LR4 use four electro-absorption modulator DFB lasers. Different lasers for the two PICs are required since the four wavelengths at 100 Gig, also around 1300nm, are more tightly spaced: 5nm versus 20nm. "They are much closer together than the 40 Gig version,” says Andy Weirich, OneChip Photonics' vice president of product line management.
Another consequence of the wider wavelength spacings is that the 40 Gig transmitter uses four discrete lasers. “Because the 40 Gig wavelengths are much further apart, putting all the lasers on the one die is problematic," says Weirich. The 40GBASE-LR4 design thus uses five indium phosphide components: four lasers and the AWG, while the 40GBASE-LR4 receiver and the two 100GBASE-LR4 devices are all monolithic PICs.
Both LR4 transmitter designs also include monitor photo-diodes for laser control
Lower size and cost
OneChip says the resulting PICs are tiny, measuring less than 3mm in length. “We think the PICs will enable the packaging of LR4 in a QSFP,” says Weirich. 40GBASE-LR4 products already exists in the QSFP form factor but the 100GBASE-LR4 uses a CFP module.
The startup expects module makers to use its receiver chips once they become available rather than wait for the receiver-transmitter PIC pair. "Reducing the size of one half the solution is possibly good enough to fit the whole hybrid design - the PIC for the receive and discretes for the transmit - into a QSFP,” says Weirich.
The PICs are expected to reduce significantly the cost of LR4 modules. "I think we can cut the price by half or better,” says Weirich. “Right now the LR4 is far too expensive to be used for data centre interconnect.” OneChip expects its LR4 PICs to be cost-competitive with the 2km reach 10x10 MSA interface.
Meanwhile, short-reach 40 and 100 Gig interfaces use VCSEL technology and multi-mode fibre to address 100m reach requirements. In larger data centres this reach is limiting. Extended reach - 300-400m - multimode interfaces have emerged but so far these are at 40 Gig only.
"[Data centre operators] are saying that they are having to significantly bend out of shape their data centre architecture to accommodate even 300m reaches,” says Weirich. “They really want more than that.”
OneChip believes interfaces distances of 200m-2km is underserved and it is this market opportunity that it is seeking to address with its LR4 designs.
Roadmap
Will OneChip integrate the design further to product a single PIC LR4 transceiver?
"It can be put into one chip but it is not clear that there is an economic advantage,” says Weirich. Indeed one PIC might even be more costly than the two-PIC chipset.
Another factor is that at 100 Gig, the 25Gbps electronics present a considerable signal integrity design challenge. “It is very important to keep the electronics very close to the photo-detectors and the modulators,” he says. “That becomes more difficult if you put it all on the one chip.” The fabrication yield of a larger single PIC would also be reduced, impacting cost.
OneChip, meanwhile, has started limited production of its PON optical network unit (ONU) transceivers based on its EPON and GPON PICs. The company's EPON transceivers are becoming generally available while the GPON transceivers are due in two months’ time.
The company has yet to decide whether it will make its own LR4 optical modules. For now OneChip is solely an LR4 component supplier.
Further reading:
See OFC/ NFOEC 2012 highlights, the Kotura story in the Optical Engines section
The CFP4 optical module to enable Terabit blades
Source: Gazettabyte, Xilinx
The CFP2 is about half the size of the CFP while the CFP4 is half the size of the CFP2. The CFP4 is slightly wider and longer than the QSFP.
The two CFP modules will use a 4x25Gbps electrical interface, doing away with the need for a 10x10Gbps to 4x25Gbps gearbox IC used for current CFP 100GBASE-LR4 and -ER4 interfaces. The CFP2 and CFP4 are also defined for 40 Gigabit Ethernet use.
The CFP's maximum power rating is 32W, the CFP2 12W and the CFP4 5W. But vendors that put eight CFP2 or 16 CFP4s on a blade still want to meet the 60W total power budget.
Getting close: Four CFP modules deliver slightly less bandwidth than 48 SFP+ modules: 4x100Gbps versus 480Gbps. The four also consume more power - 60w versus 48W. Moving to the CFP2 module will double the blade's bandwidth without consuming more power while the CFP4 will do the same again. a blade with 16 CFP4 modules promises 1.6Tbps while requiring 60W. Source: Xilinx
The first CFP2 modules are expected this year - there could be vendor announcements as early as the upcoming OFC/NFOEC 2012 show to be held in LA in the first week in March. The first CFP4 products are expected in 2013.
Further reading
The CFP MSA presentation: CFP MSA 100G roadmap and applications
Next-gen 100 Gigabit short reach optics starts to take shape
The latest options for 100 Gigabit-per-second (Gbps) interfaces are beginning to take shape following a meeting of the IEEE 802.3 Next Generation 100Gb/s Optical Ethernet Study Group in November.
The interface options being discussed include:
- A parallel multi-mode fibre using a VCSEL with a reach of 50m to 70m. An active optical cable version with a 30m reach, limited by the desired cable length rather than the technology, using silicon photonics or a VCSEL has also been proposed.
- A parallel single-mode fibre using a 1310nm electro-absorption modulated laser (EML) or silicon photonics with a range of 50m to 1000m+.
- A duplex single-mode fiber, using wavelength division multiplexing (WDM) or pulse-width modulation (PAM), an EML or silicon photonics for a 2km reach.
“I think in the end all will be adopted,” says Marek Tlalka, director of marketing at Luxtera. "Users will be able to choose what is most economical."
Jon Anderson, director of technology programme at Opnext, stresses however that these are proposals.
"No decisions were reached by the Study Group on any of these proposals," he says. “The Study Group is only working towards defining objectives for a next-gen 100 Gigabit Ethernet Optics project.” Agreement on technical solutions is outside the scope of the Study Group.
Anderson says there is a general agreement to define a 4x25Gbps multi-mode fibre optical interface. But the issues of reach and multi-mode fibre type (OM3, OM4) are still being studied.
“The Study Group has not reached any agreement on whether a 100GE short reach single-mode objective should be pursued," says Anderson. “Discussion at this point are on reach, power consumption and relative cost of possible solutions with respect to (the 10km) 100GBASE-LR4."
CyOptics gets $50m worth of new investors and funding
“Volume production scale is very important to having a successful business”
Ed Coringrato, CyOptics
The $50m investment in CyOptics has two elements: the amount paid by new investors in CyOptics to replace existing ones and funding for the company.
“This is different from the years-ago, traditional funding round but not all that different from what is more and more taking place,” says Ed Coringrato, CEO of CyOptics. “Fifty million is a big number but it is a ‘primary/ secondary’: the secondary is tendering out current investors that are choosing to exit, while the primary is what people think of as a traditional investment.” CyOptics has not detailed how the $50m is split between the two.
The funding is needed to bolster the company’s working capital, says Coringrato, despite CyOptics achieving over $100m in revenues in 2010. The money is required because of growth, he says: inventories the company holds are growing, there is more cash outstanding and the company’s payments are also rising.
There is also a need to invest in the company. “For the first time in a long time we are starting to make significant capital investments in our business,” says Coringrato. “We are ramping the fab, the packaging capability, and the assembly and test.”
The company is investing in R&D. At the moment 11 percent of its revenue is invested in R&D and the company wants to approach 13 percent. “That is a challenge in our industry – the investment in R&D is pretty significant,” says Coringrato. “If we are to continue to be significant and have leading-edge products, we must continue to make that investment.”
Manufacturing
CyOptics acquired Triquint Semiconductor’s optoelectronics operations in 2005, and before that Triquint had bought the optoelectronics operations of Agere Systems. This resulted in CyOptics inheriting automated manufacturing facilities and as a result it never felt the need to move manufacturing to the Far East to achieve cost benefits. CyOptics does use some contract manufacturing but its high-end products are made in-house.
“We have been focussed on automated production, cycle-time reduction and yield improvement,” says Coringrato. “The capital investment is to replicate what we have, adding more machines to get more output.”
Markets
CyOptics supplies fibre-to-the-x (FTTx) components to transmit optical subassembly (TOSA) and receive optical subassembly (ROSA) makers, optical transceiver players and board manufacturers. FTTx is an important market for CyOptics as it is a volume driver. “Volume production scale is very important to having a successful business,” says Coringrato.
The company also supplies 2.5 and 10 Gigabit-per-second (Gbps) TOSAs and ROSAs for XFP and SFP pluggable modules for the metro. “We want to play at the higher end as well as that is the where the growth opportunities are and the healthier margins,” says Coringrato.
CyOptics is also active in what it calls high-end product areas.
One area is as a supplier of components for the US defence industry. CyOptics entered the defence market in 2005. “These are custom products designed for specific applications,” says Stefan Rochus, vice president of marketing and business development. These include custom chip fabrication and packaging undertaking for defence contractors that supply the US Department of Defense. “When you look around there are not many companies that can do that,” says Rochus. One example CyOptics cites is a 1480nm pump-laser, part of a fibre-optic gyroscope for use in a satellite.
“We are shipping 40Gbps and 100Gbps coherent receivers into the PM-QPSK market”
Stefan Rochus, CyOptics
The defence market may require long development cycles but CyOptics believes that in the next few years several of its products could lead to reasonable volumes and a better average selling price than telecom components.
Another high-end product segment CyOptics is pursuing is photonic integrated circuits (PICs) using the company's indium-phosphide and planar lightwave circuit expertise.
Rochus says the company has several PIC developments including 10x10Gbps TOSAs and ROSAs as well as emerging 40GBASE-LR4 and coherent detection designs. “We are shipping 40Gbps and 100Gbps coherent receivers into the PM-QPSK market,” says Rochus.
CyOptics’ product portfolio is a good balance between high-volume and high average selling price components, says Rochus.
10x10 MSA
CyOptics is part of the recent 10X10 MSA, the 100Gbps multi-source agreement that includes Google and Brocade. “There is a follow-up high density 10x10Gbps MSA and we will be a member of this as well,” says Rochus. “This [10x10G design] is for short reach, up to 2km, but we are also shipping product for DWDM for an Nx10Gbps TOSA/ROSA solution.”
Why is CyOptics supporting the Google-backed 10x10Gbps MSA?
“The IEEE has only standardised the 100GBASE-SR10 which is 100m and the 100GBASE-LR4 which is 10km, there is a gap in the middle for [a] 2km [interface] which the MSA tries to solve,” says Rochus. “This is particularly important for the larger data centres.”
Rochus claims the 10x10Gbps design is the cheapest solution and that the volumes that will result from growth in the 10 Gigabit PON market will further reduce the component costs used for the interface. Furthermore the interface will be lower power.
That said, CyOptics is backing both interface styles, selling TOSAs and ROSAs for the 10x10Gbps interface and lasers for the 4x25Gbps-styled 100 Gigabit interfaces.
What next?
“The bigger we can get in terms of volume and revenue, the better our financials,” says Coringrato. “Potentially CyOptics is not only attractive for our preferred path, which is an IPO offering at the right time, but also I think it won't discourage others from being interested in us.”
Further reading
Q&A with Rafik Ward - Part 1
"This is probably the strongest growth we have seen since the last bubble of 1999-2000." Rafik Ward, Finisar
Q: How would you summarise the current state of the industry?
A: It’s a pretty fun time to be in the optical component business, and it’s some time since we last said that.
We are at an interesting inflexion point. In the past few years there has been a lot of emphasis on the migration from 1 to 2.5 Gig to 10 Gig. The [pluggable module] form factors for these speeds have been known, and involved executing on SFP, SFP+ and XFPs.
But in the last year there has been a significant breakthrough; now a lot of the discussion with customers are around 40 and 100 Gig, around form factors like QSFP and CFP - new form factors we haven’t discussed before, around new ways to handle data traffic at these data rates, and new schemes like coherent modulation.
It’s a very exciting time. Every new jump is challenging but this jump is particularly challenging in terms of what it takes to develop some of these modules.
From a business perspective, certainly at Finisar, this is probably the strongest growth we have seen since the last bubble of 1999-2000. It’s not equal to what it was then and I don’t think any of us believes it will be. But certainly the last five quarters has been the strongest growth we’ve seen in a decade.
What is this growth due to?
There are several factors.
There was a significant reduction in spending at the end of 2008 and part of 2009 where end users did not keep up with their networking demands. Due to the global financial crisis, they [service providers] significantly cut capex so some catch-up has been occurring. Keep in mind that during the global financial crisis, based on every metric we’ve seen, the rate of bandwidth growth has been unfazed.
From a Finisar perspective, we are well positioned in several markets. The WSS [wavelength-selective switch] ROADM market has been growing at a steady clip while other markets are growing quite significantly – at 10 Gig, 40 Gig and even now 100 Gig. The last point is that, based on all the metrics we’ve seen, we are picking up market share.
Your job title is very clear but can you explain what you do?
I love my job because no two days are the same. I come in and have certain things I expect to happen and get done yet it rarely shapes out how I envisaged it.
There are really three elements to my job. Product management is the significant majority of where I focus my efforts. It’s a broad role – we are very focussed on the products and on the core business to win market share. There is a pretty heavy execution focus in product management but there is also a strategic element as well.
The second element of my job is what we call strategic marketing. We spend time understanding new, potential markets where we as Finisar can use our core competencies, and a lot of the things we’ve built, to go after. This is not in line with existing markets but adjacent ones: Are there opportunities for optical transceivers in things like military and consumer applications?
One of the things I’m convinced of is that, as the price of optical components continues to come down, new markets will emerge. Some of those markets we may not even know today, and that is what we are finding. That’s a pretty interesting part of my job but candidly I spend quite a bit less time on it [strategic marketing] than product management.
The third area is corporate communications: talking to media and analysts, press releases, the website and blog, and trade shows.
"40Gbps DPSK and DQPSK compete with each other, while for 40 Gig coherent its biggest competitor isn’t DPSK and DQPSK but 100 Gig."
Some questions on markets and technology developments.
Is it becoming clearer how the various 40Gbps line side optics – DPSK, DQPSK and coherent – are going to play out?
The situation is becoming clearer but that doesn’t mean it is easier to explain.
The market is composed of customers and end users that will use all of the above modulation formats. When we talk to customers, every one has picked one, two or sometimes all three modulation formats. It is very hard to point to any trend in terms of picks, it is more on a case-by-case basis. Customers are, like us at the component level, very passionate about the modulation format that they have chosen and will have a variety of very good reasons why a particular modulation format makes sense.
Unlike certain markets where you see a level of convergence, I don’t think that there will be true convergence at 40 Gbps. Coherent – DP-QPSK - is a very powerful technology but the biggest challenge 40 Gig has with DP-QPSK is that you have the same modulation format at 100 Gig.
The more I look at the market, 40Gbps DPSK and DQPSK compete with each other, while for 40 Gig coherent its biggest competitor isn’t DPSK and DQPSK but 100 Gig.
Finisar has been quiet about its 100 Gig line side plans, what is its position?
We view these markets - 40 and 100 Gig line side – as potentially very large markets at the optical component level. Despite that fact that there are some customers that are doing vertical integrated solutions, we still see these markets as large ones. It would be foolish for us not to look at these markets very carefully. That is probably all I would say on the topic right now.
"Photonic integration is important and it becomes even more important as data rates increase."
Finisar has come out with an ‘optical engine’, a [240Gbps] parallel optics product. Why now?
This is a very exciting part of our business. We’ve been looking for some time at the future challenges we expect to see in networking equipment. If you look at fibre optics today, they are used on the front panel of equipment. Typically it is pluggable optics, sometimes it is fixed, but the intent is that the optics is the interface that brings data into and out of a chassis.
People have been using parallel optics within chassis – for backplane and other applications – but it has been niche. The reason it’s niche is that the need hasn’t been compelling for intra-chassis applications. We believe that need will change in the next decade. Parallel optics intra-chassis will be needed just to be able to drive the amount of bandwidth required from one printed circuit board to another or even from one chip to another.
The applications driving this right now are the very largest supercomputers and the very largest core routers. So it is a market focussed on the extreme high-end but what is the extreme high-end today will be mainstream a few years from now. You will see these things in mainstream servers, routers and switches etc.
Photonic integration – what’s happening here?
Photonic integration is something that the industry has been working on for several years in different forms; it continues to chug on in the background but that is not to understate its importance.
For vendors like Finisar, photonic integration is important and it becomes even more important as data rates increase. What we are seeing is that a lot of emerging standards are based around multiple lasers within a module. Examples are the 40GBASE-LR4 and the 100GBASE-LR4 (10km reach) standards, where you need four lasers and four photo-detectors and the corresponding mux-demux optics to make that work.
The higher the number of lasers required inside a given module, and the more complexity you see, the more room you have to cost-reduce with photonic integration.
Google and the optical component industry
According to a report by Pauline Rigby, Google wants something in between two existing IEEE interface standards. The 100GBase-SR10, which has 10 parallel channels and a 125m span, has too short a reach for Google.
“What is good for an 800-pound gorilla is not necessarily good for the industry. It [Google] should have been at the table when the IEEE was working on the standard."
Daryl Inniss, practice leader, components, Ovum
The second interface, the 100GBase-LR4, uses four channels that are multiplexed onto a single fibre and has a 10km reach. The issue here is that Google doesn’t need a 10km reach and while a single fibre is better than the multi-mode fibre based SR10, the interface is costly with its “gearbox” IC that translates between 10 lanes of 10Gbps and four lanes each at 25Gbps. Both IEEE interfaces are also implemented using a CFP form factor which is bulky.
What Google wants
Google wants optical component vendors to develop a new 100 Gigabit Ethernet multi-source agreement (MSA) that is based on a single-mode interface with a 2km reach, reports Rigby. Such a design would use a ten-channel laser array whose output is multiplexed onto a fibre, a similar laser array-multiplexer arrangement that has already been developed by Santur. Using such a part, the new interface could be developed quickly and cheaply, says Google.
The proposed interface clearly has merits and Google, an important force with an appetite for optics, makes some valid points. But the industry is developing 4x25Gbps interfaces and while such interfaces may be challenging, no-one doubts they will come.
Google’s next moves
Google has a history of being contrarian if it believes it best serves its business. The way the internet giant designs data centres is one example, using massive numbers of cheap servers arranged in a fault-tolerant architecture.
But there is only so much it can do in-house and developing a new optical interface will require help from optical component players.
Google has the financial muscle to hire an optical component firm to engineer and manufacture a custom interface. A recent example of such a partnership is IBM's work with Avago Technologies to develop board-level optics – or an optical engine – for use within IBM’s POWER7 supercomputer systems.
According to Karen Liu, vice president, components and video technologies at market research firm Ovum, once such an interface is developed, Google could allow others to buy it to help reduce its price. “Remember the Lucent form factor which became a de facto standard but wasn’t originally intended to be?” says Liu. “This approach could work.”
Taking a longer term view, Google could also invest in optical component start-ups. The return may take years and as the experience of the last decade has shown, optical components is a risky business. But Google could encourage a supply of novel, leading-edge technologies over the next decade.
The optical component industry is right to push back with regard Google’s request for a new 100 Gigabit Ethernet MSA, as Finisar has done. While Google may be an important player that can drive interface requirements, many players have helped frame the IEEE 100Gbps Ethernet standards work. In the last decade the optical industry has also seen other giant firms try to drive the industry only to eventually exit.
“The industry needs to move on,” says Daryl Inniss, practice leader, components at Ovum. “What is good for an 800-pound gorilla is not necessarily good for the industry.” Inniss also suggests a simple and effective way Google could have influenced the 100 Gigabit Ethernet MSA work: “It [Google] should have been at the table when the IEEE was working on the standard."