Industry underestimating 25 Gigabit parallel optics challenge
Ten Gigabit-based parallel optics is set to dominate the marketplace for several years to come. So claims datacom module specialist, Avago Technologies.
"One customer told us it has to keep the interface speed below 20Gbps due to the cost of the SerDes"
Sharon Hall, Avago
"People are underestimating what is going to be involved in doing 25 Gigabit [channels]," says Sharon Hall, product line manager for embedded optics at Avago Technologies. "Ten Gigabit is going to last quite a bit longer because of the price point it can provide."
Eventually 25 Gig-based parallel optics, with its lower lane count, will be cheaper than 10 Gigabit - but is will take several years. One challenge is the cost of 25 Gigabit-per-second (Gbps) electrical interfaces, due to the large relative size of the circuitry. One customer told Avago that it has to keep the interface speed below 20Gbps for now due to the cost of the serial/ deserialiser (SerDes).
Avago has announced that its 120 Gigabit aggregate bandwidth (12x10Gbps) MiniPod and CXP parallel optics products are now in volume production. The company first detailed the MiniPod and CXP technologies in late 2010 yet many equipment makers are still to launch their first designs.
The CXP is a pluggable optical transceiver while the MiniPod is Avago's packaged optical engine used for embedded designs. The 22x18mm MiniPod is based on Avago's 8x8mm MicroPod optical engine but uses a 9x9 electrical MegArray connector with its more relaxed pitch.
Equipment makers face a non-trivial decision as to whether to adopt copper or optical interfaces for their platform designs. "This is a major design decision with a lot of customers going back and forth deciding which way to go," says Hall. "They might do a mix with some short connections staying copper but if they need 10 Gig at anything longer than a few meters then they are going to go optical."
Having chosen parallel optics, the style of form factor - pluggable or embedded - is largely based on the interface density required. "Certain customers prefer field pluggability [of CXP] with its pay-as-you-go and ease of installation features, but are limited on port density due to the number of CXP transceivers that can physically fit on a 19 inch board," says Hall.
Up to 14 CXPs can fit onto a 19-inch board. In contrast, some 50-100 transmit and receive MiniPod pairs can fit on the 19-inch board. "You have the whole board space to work with," she says. The embedded optics sit closer to the board's ASICs, shortening the electrical path and solving signal integrity issues that can arise using edge-mounted pluggables. Thermal management - not having all the pluggable optics at the card edge furthest from the fans - is also simplified using embedded optics.
Generally, connections to data centre top-of-rack switches and between chassis use the pluggable CXP while internal backplane and mid-plane designs use the MiniPod. The CXP is also used by core switches and routers; Alcatel-Lucent's recently announced 7950 core router has a four-port CXP-based card. But Avago stresses that there are no hard rules: It has customers that have chosen the CXP and others the MiniPod for the same class of platform.
Source: Gazettabyte
25 Gigabit parallel optics
Finisar recently demonstrated its board mounted optical assembly that it says will support channel speeds of 10, 14, 25 and 28Gbps, while silicon photonics vendors Luxtera and Kotura have announced 4x25Gbps optical engines. OneChip Photonics has announced photonic integrated circuits for the 4x25Gbps, 100GBase-LR4 10km standard that will also address short and mid-reach applications
Avago has yet to make an announcement regarding higher speed parallel optics. "It is just a matter of time," says Hall. "We have done a demonstration of our 25Gbps VCSEL in an SFP+ package over a year ago, and we are developing parallel optics 25Gbps solutions."
But 25Gbps will take time before it gets to volume production, says Hall: "It is going to be a long, long design cycle for system companies - doing 25Gbps on their boards and their systems is a completely new design."
Supercomputers and system mid-plane and backplane applications could happen a lot earlier than 4x25GbE applications. "Some customers are interested in getting 4x25Gbps samples in the 2013 timeframe," says Hall. "But we expect that volume is going to take at least another year from that."
Meanwhile, Avago says it has already shipped 600,000 MicroPods which has been generally available for over a year.
Altera optical FPGA in 100 Gigabit Ethernet traffic demo
Altera is demonstrating its optical FPGA at OFC/NFOEC, being held in Los Angeles this week. The FPGA, coupled to parallel optical interfaces, is being used to send and receive 100 Gigabit Ethernet packets of various sizes.
The technology demonstrator comprises an Altera Stratix IV FPGA with 28, 11.3Gbps electrical transceivers coupled to two Avago Technologies' MicroPod optical modules.
"FPGAs are now being used for full system level solutions"
Kevin Cackovic, Altera
The MicroPods - a 12x10Gbps transmitter and a 12x10Gbps optical transceiver - are co-packaged with the FPGA. "All the interconnect between the serdes and the optics are on the package, not on the board," says Steve Sharp, marketing program manager, fiber optic products division at Avago. Such a design benefits signal integrity and power consumption, he says: "It opens up a different world for FPGA users, and for system integration for optic users."
Both Altera and Avago stress that the optical FPGA has been designed deliberately using proven technologies. "We wanted to focus on demonstrating the integration of the optics, not pushing either of the process technologies to the absolute edge," says Sharp.
The nature of FPGA designs has changed in recent years, says Kevin Cackovic, senior strategic marketing manager of Altera's transmission business unit. Many designs no longer use FPGAs solely to interface application-specific standard products to ASICs, or as a co-processor. "FPGAs are now being used for full system level solutions, things like a framer or MAC technology, forward error correction at very high rates, mapper engines, packet processing and traffic management," he says.
Having its FPGAs in such designs has highlighted for Altera current and upcoming system bottlenecks. "This is what is driving our interest in looking at this technology and what is possible integrating the optics into the FPGA," says Cackovic. Applications requiring the higher bandwidth and the greater reach of optical - rack-to-rack rather than chip-to-module - include next-generation video, cloud computing and 3D gaming, he says.
Altera has still to announce its product plans regarding the optical FPGA dsign. Meanwhile Avago says it is looking at higher-speed versions of MicroPod.
"The request for higher line rates is obviously there," says Sharp. "Whether it goes all the way to 28 [Gigabit] or one of the steps in-between, we are not sure yet."
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.