Rockley Photonics showcases its in-packaged design at OFC
The packaged design includes Rockley's own 2 billion transistor layer 3 router chip, and its silicon photonics-based optical transceivers. The layer 3 router chip, described as a terabit device, also includes mixed-signal circuits needed for the optical transceevers' transmit and receive paths.
Source: Rockley Photonics (annotated by Gazettabyte).
Rockley says it is using 500m-reach PSM4 transceivers for the design and that while a dozen ribbon cables are shown, this does not mean there are 12 100-gigabit PSM4 transceivers. The company is not saying what the total optical input-output is.
Source: Rockley Photonics (annotated by Gazettabyte).
The company has said it is not looking to enter the marketplace as a switch chip player competing with the likes of Broadcom, Intel, Cavium, Barefoot Networks and Innovium. To develop such a device and remain competitive requires considerable investment and that is not Rockley's focus. Instead, it is using its router chip as a demonstrator to show the marketplace what can be done and that the technology works.
When asked what progress Rockley is making showcasing its technology, its CEO Andrew Rickman said: “It is going very well but nothing we can say publicly."
The switch chip makers continue to use electrical interfaces for their state-of-the-art switches which have a capacity of 12.8 terabits. It still remains to be seen which generation of switch chip will finally adopt in-packaged optics and whether on-board optics designs such as COBO will be adopted first.
For the full interview with CEO Andrew Rickman, click here.
Rockley Photonics eyes multiple markets
Andrew Rickman, founder and CEO of silicon photonics start-up, Rockley Photonics, discusses the new joint venture with Hengtong Optic-Electric, the benefits of the company’s micron-wide optical waveguides and why the timing is right for silicon photonics.
Andrew Rickman
The joint venture between Rockley Photonics and Chinese firm Hengtong Optic-Electric is the first announced example of Rockley’s business branching out.
The start-up’s focus has been to apply its silicon photonics know-how to data-centre applications. In particular, Rockley has developed an Opto-ASIC package that combines optical transceiver technology with its own switch chip design. Now it is using the transceiver technology for its joint venture.
“It was logical for us to carve out the pieces generated for the Opto-ASIC and additionally commercialise them in a standard transceiver format,” says Andrew Rickman, Rockley’s CEO. “That is what the joint venture is all about.”
Rockley is not stopping there. Rickman describes the start-up as a platform business, building silicon photonics and electronics chipsets for particular applications including markets other than telecom and datacom.
Joint venture
Hengtong and Rockley have set up the $42 million joint venture to make and sell optical transceivers.
Known for its optical fibre cables, Hengtong is also a maker of optical transceivers and owns 75.1 percent of the new joint venture. Rockley gains the remaining 24.9 percent share in return for giving Hengtong its 100-gigabit QSFP transceiver designs. The joint venture also becomes a customer of Rockley’s, buying its silicon photonics and electronics chips to make the QSFP modules.
“Hengtong is one of the world’s largest optical fibre cable manufacturers, is listed on the Shanghai stock market, and sells extensively in China and elsewhere into the data centre market,” says Rickman. “It is a great conduit, a great sales channel into these customers.”
The joint venture will make three 100-gigabit QSFP-based products: a PSM4 and a CWDM4 pluggable module and an active optical cable. Rickman expects the joint venture to make other module designs and points out that Rockley participates in the IEEE standards work for 400 gigabits and is one of the co-founders of the 400-gigabit CWDM8 MSA.
Rockley cites several reasons why the deal with Hengtong makes sense. First, a large part of the bill of materials used for active optical cables is the fibre itself, something which the vertically integrated Hengtong can provide.
China also has a ‘Made in China 2025’ initiative that encourages buying home-made optical modules. Teaming with Hengtong means Rockley can sell to the Chinese telecom operators and internet content players.
In addition, Hengtong is already doing substantial business with all of the global data centres as a cable, patch panel and connector supplier, says Rickman:“So it is an immediate sales channel into these companies without having to break into these businesses as a qualified supplier afresh.”
A huge amount of learning happened and then what Rockley represented was the opportunity to start all over again with a clean sheet of paper but with all that experience
Bigger is Best?
At the recent SPIE Photonics West conference held in San Francisco, Rickman gave a presentation entitled Silicon Photonics: Bigger is Better. His talk outlined the advantages of Rockley’s use of three-micron-wide optical waveguides, bucking the industry trend of using relatively advanced CMOS processes to make silicon photonics components.
Rickman describes as seductive the idea of using 45nm CMOS for optical waveguides.“These things exist and work but people are thinking of them in the same physics that have driven microelectronics,” he says. Moving to ever-smaller feature sizes may have driven Moore’s Law but using waveguide dimensions that are smaller than the wavelength of light makes things trickier.
To make his point, he plots the effective index of a waveguide against its size in microns. The effective index is a unitless measure - a ratio of a phase delay in a unit length of a waveguide relative to the phase delay in a vacuum. “Once you get below one micron, you get a waveguide that is highly polarisation-dependent and just a small variation in the size of the waveguide has a huge variation in the effective index,” says Rickman.
Such variations translate to inaccuracies in the operating wavelength. This impacts the accuracy of circuits, for example, arrayed-waveguide gratings built using waveguides to multiplex and demultiplex light for wavelength-division multiplexing (WDM).
“Above one micron is where you want to operate, where you can manufacture with a few percent variation in the width and height of a waveguide,” says Rickman.“But the minute you go below one micron, in order to hit the wavelength registration that you need for WDM, you have got to control the [waveguide’s] film thickness and line thickness to fractions of a percent.” A level of accuracy that the semiconductor industry cannot match, he says.
A 100GHz WDM channel equates to 0.8nm when expressed using a wavelength scale. “In our technology, you can easily get a wavelength registration on a WDM grid of less than 0.1nm,” says Rickman. “Exactly the same manufacturing technology applied to smaller waveguides is 25 times worse - the variation is 2.5nm.”
Moreover, WDM technology is becoming increasingly important in the data centre. The 100-gigabit PSM4 uses a single wavelength, the CWDM4 uses four, while the newer CWDM8 MSA for 400 gigabit uses eight wavelengths. “In telecom, 90-plus wavelengths can be used; the same thing will come to pass in the years to come in data centre devices,” he says.
Rockley also claims it has a compact modulator that is 50 times smaller than competing modulators despite them being implemented using nanometer feature sizes.
We set out to generate a platform that would be pervasive across communications, new forms of advanced computing, optical signal processing and a whole range of sensor applications
Opto-ASIC reference design
Rockley’s first platform technology example is its Opto-ASIC reference design. The design integrates silicon photonics-based transceivers with an in-house 2 billion transistor switch chip all in one package. Rockley demonstrated the technology at OFC 2017.
“If you look around, this is something the industry says is going to happen but there isn't a single practical instantiation of it,” says Rickman who points out that, like the semiconductor industry, very often a reference design needs to be built to demonstrate the technology to customers.“So we built a complete reference design - it is called Topanga - an optical-packaged switch solution,” he says.
Despite developing a terabyte-class packet processor, Rockley does not intend to compete with the established switch-chip players. The investment needed to produce a leading edge device and remain relevant is simply too great, he says.
Rockley has demonstrated its in-package design to relevant companies. “It is going very well but nothing we can say publicly,” says Rickman.
New Markets
Rockley is also pursuing opportunities beyond telecom and datacom.
“We set out to generate a platform that would be pervasive across communications, new forms of advanced computing, optical signal processing and a whole range of sensor applications,” says Rickman.
Using silicon photonics for sensors is generating a lot of interest. “We see these markets starting to emerge and they are larger than the data centre and communications markets,” he says. “A lot of these things are not in the public domain so it is very difficult to report on.”
Moreover, the company’s believes its technology gives it an advantage for such applications. “When we look across the other application areas, we don’t see the small waveguide platforms being able to compete,” says Rickman. Such applications can use relatively high power levels that exceed what the smaller waveguides can handle.
Rockley is sequencing the markets it will address. “We’ve chosen an approach where we have looked at the best match of the platform to the best opportunities and put them in an order that makes sense,” says Rickman.
Rockley Photonics represent Rickman’s third effort to bring silicon photonics to the marketplace.Bookham Technology, the first company he founded, build different prototypes in several different areas but the market wasn't ready. In 2005 he joined start-up Kotura as a board member. “A huge amount of learning happened and then what Rockley represented was the opportunity to start all over again with a clean sheet of paper but with all that experience,” says Rickman.
Back in 2013, Rockley saw certain opportunities for its platform approach and what has happened since is that their maturity and relevance has increased dramatically.
“Like all things it is always down to timing,” says Rickman. “The market is vastly bigger and much more ready than it was in the Bookham days.”
Professor Graham Reed: The calm before the storm
Silicon photonics luminaries series
Interview 3: Professor Graham Reed
Despite a half-century track record driving technology, electronics is increasingly calling upon optics for help. “It seems to me that this is a marriage that is really going to define the future,” says Graham Reed, professor of silicon photonics at the University of Southampton’s Optoelectronics Research Centre.
The optics alongside the electronics does not have to be silicon photonics, he says, but silicon as a photonics technology is attractive for several reasons.
“What makes silicon photonics interesting is its promise to enable low-cost manufacturing, an important requirement for emerging consumer applications,” says Reed. And being silicon-based, it is much more compatible than other photonics technologies. “It probably means silicon photonics is going to win out,” he says.
From Surrey to Southampton
Reed has been active in silicon photonics for over 25 years. As an academic at the University of Surrey, his first Ph.D. student was Andrew Rickman, who went on to found Bookham Technology and is now CEO of Rockley Photonics.
Rickman undertook the study of basic optical waveguide structures using silicon. “The first data we got, the waveguide losses were very high, 20 to 30dB per centimetre,” says Reed. “Within a year, we got the losses down to below 1dB per centimetre; that makes it viable.”
The research then broadened to include silicon modulators, a research topic Reed continues to this day.
Everything about silicon photonics is about low cost
The optical modulator is silicon photonics biggest achievement to date, argues Reed. “We were working on modulators in 1991 that worked at 20 megahertz,” he says. “Intel’s Mario Paniccia ribbed me when they got [a modulator] to 1 gigahertz.”
The Surrey group was not focussing on telecom when they started. “I never believed in the early 1990s that these things were going to go as fast as they became,” says Reed. Partly that was because the early work used much larger waveguides and to increase speed, the dimensions need to shrink.
In 2012, Reed and a dozen colleagues moved from the University of Surrey to the University of Southampton. Several factors led to the move. The University of Southampton was interested in the team, given its reputation and the rising importance of silicon photonics, while Reed was keen to make use of the university’s new on-site fabrication plant, which he describes as the best university fab in the UK and probably Europe.
“We were increasing frustrated with the fab facilities around the world,” says Reed. The team used multi-project wafers where companies and institutions have their circuits made on a shared wafer. However, such multi-project wafers have a lower run priority.
“Foundries do a good job but they often take much longer to deliver [the designs] than they aim,” says Reed. Worst case, it can take over three years to receive the chip design back. Given a project cycle typically lasts three years, this is a non-starter, he says: “Having a fab that you have a lot of control over is a big attraction”.
Research focus
Reed’s group is regularly approached by companies from all over the world. But it wasn't always like that. In the 1990s, getting funding to research silicon photonics was a challenge, he says.
The companies now contacting Reed’s group are either in the field and have a difficulty, or they want to enter the marketplace. “They want particular work done or a particular device worked upon,” he says.
Intel is one company that worked with Reed when they started their silicon photonics programme some dozen years ago.
Reed’s group’s research covers the development of individual optical components as well as systems. Much of the work is focussed on telecom and datacom, given that is where silicon photonics is most established, but the group is also conducting work using silicon photonics for longer wavelengths - 2 to 18 microns - known as the mid infra-red region.
Mid infra-red is an emerging field, says Reed: “People have seen the success of existing silicon photonics and are applying it to longer wavelengths.”
Such wavelengths are suited for sensing applications. “A lot of nasties - chemicals you’d want to sense - have characteristic absorption lines in this longer wavelength range,” he says.
Things also become easier at the longer wavelengths because the dimensions of the silicon features are more relaxed. However, additional materials are required that are transparent at these longer wavelengths, and these platforms all need developing. “Longer wavelengths equate to bigger waveguides; what gets more difficult are the sources and the detectors,” says Reed.
A third research activity his group is tackling is ongoing silicon photonics challenges such as wafer-scale testing, passive alignment, lowering power consumption and thermal stability issues.
Optical device work
Reed cites a low-channel-count multiplexer as an example of its research work on basic optical devices with the goal of helping commercialise silicon photonics.
“One of the issues in silicon photonics is to make things reliable and high yield,” says Reed. “One way to look at that is you need simplicity.”
The group has developed an angled multi-mode interference (MMI) multiplexer suited for 4 or 8 channel designs.
“It is so simple,” says Reed. The multiplexer is made in a single etch step and is based on large multi-mode waveguides that are more resilient to fabrication errors and layer thickness variations. The design is also more thermally stable than single-mode waveguides.
Another area is ring resonators - useful devices that can be used for a variety of tasks including modulation but which are sensitive to layer thickness variations as well as thermal stability issues. “If anyone is going to adopt ring resonators they need to find a way to make them athermal,” says Reed. “And they need a way to tune or trim to operate them to the resonance they need.”
Systems work
The group’s systems work addresses some of the same issues as the large systems vendors. However, the group is careful in the topics it chooses given their more modest university resources. “We are looking at more complex modulation systems but probably not for long haul communications,” says Reed.
Another research activity is looking at alternative ways to combine components. Using silicon photonics for integration in the mid infra-red range may give a new lease of life to the lab-on-a-chip concept. “People have talked about it for a long while but it hasn't really happened,” says Reed. “If you can do these things in a reliable and low-cost manner, maybe disposable chips are viable again.”
Silicon photonics challenges
Two current manufacturing challenges Reed highlights are the issues of passive alignment and wafer-scale testing.
Coupling the laser to a fibre or the silicon chip’s waveguide using passive alignment remains an ongoing challenge. “Everything about silicon photonics is about low cost,” says Reed. At present to attach a laser, it is typically turned on and aligned to the chip’s waveguide. This requires manual intervention and is time-consuming.
“The ideal scenario is to put a fibre down and it couples to the waveguide or laser and somehow you have aligned it,” he says. The challenge is the discrepancy in dimensions between the 10-micron fibre core and the waveguide, which is typically between 0.35- and 0.5-microns wide. Work is on-going to use mode converters or grating couplers such that the resulting optical loss is low enough to make passive alignment viable.
All these events are consistent with this field of technology pointing to mass markets
Wafer-scale testing remains another challenge. Grating couplers are one way designs can be tested while still on the silicon wafer. But these typically only allow the whole circuit to be tested - either it works or not - but you can’t test individual components. “If you are going to mimic the successes of electronics, you need to test more comprehensibly than that,” says Reed.
His group has developed an erasable grating that can be placed either side of a critical component to test it. These gratings can then be removed from the final circuit by using local laser annealing.
Reed expects the industry to overcome all these manufacturing challenges: “But it still means somebody has to have the brilliant idea”.
He is also somewhat surprised that there are not more silicon photonics products on the market, especially considering the huge investment in the technology made by some of the larger companies over the last decade.
He describes what is happening now as silicon photonics’ quiet period. Partly it is due to the vendors working to commercialise their technologies, partly it is the systems vendors that are developing next-generation products are evaluating the various technologies. “Until somebody jumps and that market takes off - and somebody will jump,” he says. “Then there will be ferocious activity.”
Opportunities
Reed is measured when assessing the future opportunities for the technology.
“It is not something that we strategise about - it is not what we do - but we get insights from time to time because of the people we work with and what they want,” he says. “The crucial thing is what facilitates the mass market because silicon photonics is really trying to bring photonics to the mass market.”
Reed does believe silicon photonics is disruptive: “If you look at the origins of what a disruptive technology is, it is a technology that works in one field but then it performs so well, it crosses the boundary into other areas”.
Silicon photonics was initially regarded as a short-reach technology but once the performance of its modulators started to drastically increase, the technology crossed the boundary into long-haul research, he notes. “That is the definition of a disruptive technology,” he says.
He also believes the technology has passed its tipping point. As evidence, he points to the investment made by the large companies and says it is inevitable that they will launch products: “So in that sense, the tipping point has already been and gone”.
In addition, he highlights the American Institute for Manufacturing Integrated Photonics (AIM Photonics) venture, the $610 million public and private funded initiative set up in 2015 to advance silicon photonics-based manufacturing.
“All these events are consistent with this field of technology pointing to mass markets,” says Reed. “If this was going to be indium phosphide that did that, why did not all that activity happen years ago?”
Tackling system design on a data centre scale
Interview 1: Andrew Rickman
Silicon photonics has been a recurring theme in the career of Andrew Rickman. First, as a researcher looking at the feasibility of silicon-based optical waveguides, then as founder of Bookham Technologies, and after that as a board member of silicon photonics start-up, Kotura.
Andrew Rickman
Now as CEO of start-up Rockley Photonics, his company is using silicon photonics alongside its custom ASIC and software to tackle a core problem in the data centre: how to connect more and more servers in a cost effective and scaleable way.
Origins
As a child, Rickman attended the Royal Institution Christmas Lectures given by Eric Laithwaite, a popular scientist who was also a professor of electrical engineering at Imperial College. As an undergraduate at Imperial, Rickman was reacquainted with Professor Laithwaite who kindled his interest in gyroscopes.
“I stumbled across a device called a fibre-optic gyroscope,” says Rickman. “Within that I could see people starting to use lithium niobate photonic circuits.” It was investigating the gyroscope design and how clever it was that made Rickman wonder whether the optical circuits of such a device could be made using silicon rather than exotic materials like lithium niobate.
“That is where the idea triggered, to look at the possibility of being able to make optical circuits in silicon,” he says.
If you try and force a photon into a space shorter than its wavelength, it behaves very badly
In the 1980s, few people had thought about silicon in such a context. That may seem strange today, he says, but silicon was not a promising candidate material. “It is not a direct band-gap material - it was not offering up the light source, and it did not have a big electro-optic effect like lithium niobate which was good for modulators,” he says. “And no one had demonstrated a low-loss single-mode waveguide.”
Rickman worked as a researcher at the University of Surrey’s physics department with such colleagues as Graham Reed to investigate whether the trillions of dollars invested in the manufacturing of silicon could also be used to benefit photonic circuits and in particular whether silicon could be used to make waveguides. “The fundamental thing one needed was a viable waveguide,” he says.
Rickman even wrote a paper with Richard Soref who was collaborating with the University of Surrey at the time. “Everyone would agree that Richard Soref is the founding father of the idea - the proposal of having a useful waveguide in silicon - which is the starting point,” says Rickman. It was the work at the University of Surrey, sponsored by Bookham which Rickman had by then founded, that demonstrated low-loss waveguides in silicon.
Fabrication challenges
Rickman argues that not having a background in CMOS processes has been a benefit. “I wasn’t dyed-in-the-wool-committed to CMOS-type electronics processing,” he says. “I looked upon silicon technology as a set of machine-shop processes for making things.”
Looking at CMOS processing completely afresh and designing circuits optimised for photonics yielded Bookham a great number of high-performance products, he says. In contrast, the industry’s thrust has been very much a semiconductor CMOS-focused one. “People became interested in photonics because they just naturally thought it was going to be important in silicon, to perpetuate Moore’s law,” says Rickman.
You can use the structures and much of the CMOS processes to make optical waveguides, he says, but the problem is you create small structures - sub-micron - that guide light poorly. “If you try and force a photon into a space shorter than its wavelength, it behaves very badly,” he says. “In microelectronics, an electron has got a wavelength that is one hundred times smaller that the features it is using.”
The results include light being sensitive to interface roughness and to the manufacturing tolerances - the width, hight and composition of the waveguide. “At least an order of magnitude more difficult to control that the best processes that exist,” says Rickman.
“Our [Rockley’s] waveguides are one thousand times more relaxed to produce than the competitors’ smaller ones,” he says. “From a process point of view, we don’t need the latest CMOS node, we are more a MEMS process.”
If you take control of enough of the system problem, and you are not dictated to in terms of what MSA or what standard that component must fit into, and you are not competing in this brutal transceiver market, then that is when you can optimise the utilisation of silicon photonics
Rickman stresses that small waveguides do have merits - they go round tighter bends, and their smaller-dimensioned junctions make for higher-speed components. But using very large features solves the ‘fibre connectivity problem’, and Rockley has come up with its own solutions to achieve higher-speed devices and dense designs.
“Bookham was very strong in passive optics and micro-engineered features,” says Rickman. “We have taken that experience and designed a process that has all the advantages of a smaller process - speed and compactness - as well as all the benefits of a larger technology: the multiplexing and demultiplexing for doing dense WDM, and we can make a chip that already has a connector on it.”
Playing to silicon photonics’ strengths
Rickman believes that silicon photonics is a significant technological development: “It is a paradigm shift; it is not a linear improvement”. But what is key is how silicon photonics is applied and the problem it is addressing.
To make an optical component for an interface standard or a transceiver MSA using silicon photonics, or to use it as an add-on to semiconductors - a ’band-aid” – to prolong Moore’s law, is to undersell its full potential. Instead, he recommends using silicon photonics as one element - albeit an important one - in an array of technologies to tackle system-scale issues.
“If you take control of enough of the system problem, and you are not dictated to in terms of what MSA or what standard that component must fit into, and you are not competing in this brutal transceiver market, then that is when you can optimise the utilisation of silicon photonics,” says Rickman. “And that is what we are doing.” In other words, taking control of the environment that the silicon sits in.
It [silicon photonics] is a paradigm shift; it is not a linear improvement
Rockley’s team has been structured with the view to tackle the system-scale problem of interconnecting servers in the data centre. Its team comprises computer scientists, CMOS designers - digital and analogue - and silicon photonics experts.
Knowing what can be done with the technologies and organising them allows the problems caused by the ‘exhaustion of Moore’s law’ and the input/output (I/O) issues that result to be overcome. “Not how you apply one technology to make up for the problems in another technology,” says Rickman.
The ending of Moore’s law
Moore’s law continues to deliver a doubling of transistors every two years but the associated scaling benefits like the halving of power consumed per transistor no longer apply. As a result, while Moore’s law continues to grow gate count that drives greater computation, the overall power consumption is no longer constant.
Rickman also points out that the I/O - the number of connections on and off a chip - are not doubling with transistor count. “I/O may be going from 25 gigabit to 50 gigabit using PAM–4 but there are many challenges and the technology has yet to be demonstrated,” he says.
The challenge facing the industry is that increasing the I/O rate inevitably increases power consumption. “As power consumption goes up, it also equates to cost,” says Rickman. Clearly that is unwelcome and adds cost, he says, but that is not the only issue. As power goes up, you cannot fully benefit from the doubling transistor counts, so things cannot be packed more densely.
“You are running into to the end of Moore’s law and you don’t get the benefit of reducing space and cost because you’ve got to bolt on all these other things as it is very difficult to get all these signals off-chip,” he says.
This is where tackling the system as a whole comes in. You can look at microelectronics in isolation and use silicon photonics for chip-to-chip communications across a printed circuit board to reduce the electrical losses through the copper traces. “A good thing to do,” stresses Rickman. Or you can address, as Rockley aims to do, Moore’s law and the I/O limitations within a complete system the size of the data centre that links hundred of thousands of computers. “Not the same way you’d solve an individual problem in an individual device,” says Rickman.
Rockley Photonics
Rockley Photonics has already demonstrated all the basic elements of its design. “That has gone very well,” says Rickman.
The start-up has stated its switch design uses silicon photonics for optical switching and that the company is developing an accompanying controller ASIC. It has also developed a switching protocol to run on the hardware. Rockley’s silicon photonics design performs multiplexing and demultiplexing, suggesting that dense WDM is being used as well as optical switching.
Rockley is a fabless semiconductor company and will not be building systems. Partly, it is because it is addressing the data centre and the market has evolved in a different way to telecoms. For the data centre, there are established switch vendors and white-box manufacturers. As such, Rockley will provide its chipset-based reference design, its architecture IP and the software stack for its customers. “Then, working with the customer contract manufacturer, we will implement the line cards and the fabric cards in the format that the particular customer wants,” says Rickman.
The resulting system is designed as a drop-in replacement for the large-scale data centre players’ switches they haver already deployed, yet will be cheaper, more compact and consume less power, says Rockley.
“They [the data centre operators] can scale the way they do at the moment, or they can scale with our topology,” says Rickman.
The start-up expects to finally unveil its technology by the year end.
Rockley demos a silicon photonics switch prototype
Rockley Photonics has made a prototype switch to help grow the number of servers that can be linked in a data centre. The issue with interconnection networks inside a data centre is that they do not scale linearly as more servers are added.
Dr. Andrew Rickman
“If you double the number of servers connected in a mega data centre, you don’t just double the complexity of the network, it goes up exponentially,” explains Andrew Rickman, co-founder, chairman and CEO at Rockley Photonics. “That is the problem we are addressing.”
By 2017 and 2018, it will still be possible to build the networks that large-scale data centre network operators require, says Rickman, but at an ever increasing cost and with a growing power consumption. “The basic principles of what they are doing needs to be rethought,” he says.
Network scale
Modern data centre networks must handle significant traffic flow between servers, referred to as east-west traffic. A common switching arrangement in the data centre is the leaf-spine architecture, used to interconnect thousands of servers.
A ‘leaf’ may be a top-of-rack switch that is linked to multiple server chassis on one side and larger-capacity, ‘spine’ switches on the other. The result is a switch network where each leaf is connected to all the spine switches, while each spine switch is linked to all the leaves. In the example shown, four spine switches connect to 32 leaf switches.
A leaf-spine architecture
The leaf and spine switches are built using ASICs, with the largest ICs typically having 32, 100 gigabit ports. One switch ASIC may be used in a platform but as Rickman points out, larger switches may implement multiple stages such as a three-stage Clos architecture. As a result, traffic between servers on different leaves, travelling up and down the leaf-spine, may pass through five stages or hops but possibly as many as nine.
There is no replacement performance in this area
It is the switch IC’s capacity and port count that dictates the overall size of the leaf-spine network and therefore the number of servers that can be connected. Rockley’s goal is to develop a bigger switch building block making use of silicon photonics.
“The fundamental thing to address is making bigger switching elements,” says Rickman. “That way you can keep the number of stages in the network the same but still make bigger and bigger networks.” Rockley expects its larger building-block switch will reduce the switch stages needed.
The UK start-up is not yet detailing its switch beyond saying it uses optical switching and that the company is developing a photonic integrated circuit (PIC) and a controlling ASIC.
“In the field of silicon photonics, for the same area of silicon, you can produce a larger switch; you have more capacity than you do in electronics,” says Rickman. Moreover, Rockley says that its silicon photonics-based PIC will scale with Moore’s law, with its switch's data capacity approximately doubling every two years. “Previously, the network did not scale with Moore’s law,” says Rickman.
Customers can see something is real and that it works. We are optimising all the elements of the system before taping out the fully integrated devices
Status
The company has developed a switch prototype that includes ‘silicon photonics elements’ and FPGAs. “Customers can see something is real and that it works,” says Rickman. “We are optimising all the elements of the system before taping out the fully integrated devices.” Rockley expects to have its switch in volume production in 2017.
Last year the company raised its first round of funding and said that it would undergo a further round in 2015. Rockley has not said how much it has raised or the status of the latest round. “We are well-funded and we have a very supportive group of investors,” says Rickman.
Rickman has long been involved in silicon photonics, starting out as a researcher at the University of Surrey developing silicon photonics waveguides in the early 1990s, before founding Bookham Technologies (now Oclaro). He has also been chairman of silicon photonics start-up Kotura that was acquired by Mellanox Technologies in 2013. Rickman co-founded Rockley in 2013.
“What I’ve learned about silicon photonics, and about all those electronics technologies, is how to design stuff from a process point of view to make something highly manufacturable and at the same time having the performance,” says Rickman.
There is no replacement performance in the area of data centre switching, he stresses: “The benefit of our technology is to deliver the performance, not the fact that it is cheap or [offers] average performance.”
For Part 2, Interconnection networks - an introduction, click here