NetLogic Microsystems introduces low-power 10/40/100GE PHY for next-generation data center Ethernet applications/Can the optical transceiver market sustain double-digit growth?

JULY 12, 2010 — NetLogic Microsystems Inc., maker of high-performance intelligent semiconductor solutions for next-generation Internet networks, announced its low-power 10/40/100 Gigabit Ethernet PHY for next-generation data center applications.

The new NLP1342 PHY quad-port device integrates receive equalization and transmit pre-emphasis technologies with ultra low latency to address new datacenter opportunities, says a company representative. The device supports 10Gbps SFI-to-XFI, 40Gbps XLPPI-to-XLAUI, and 100Gbps CPPI-to-CAUI modes to enable migration of data centers from 10 Gigabit to 100 Gigabit throughputs.

The exponential demand for video, IPTV, social networking, peer-to-peer, and virtualization services over the Internet has caused network bottlenecks in data centers that are serving media-rich content and services, driving the migration of servers and networking equipment in the data center from 10 Gigabit links to 40 and 100 Gigabit to support next-generation services.

Minimizing the latency in transmitting network traffic has become an important performance metric in the data center, as it can have a profound impact on the overall application performance. It is particularly important because data centers are responsible for managing mission critical and time sensitive transactions, such as financial trades, web-based services, and video content delivery. Unlike enterprise or telco infrastructure networks that transmit packets over longer distances, data center devices are usually only a few meters apart so every micro-second reduction in latency has a very large impact on the overall application performance.

The ability to drive 10/40/100 Gbps transmission over either twin-ax copper or optical cable makes NetLogic Microsystems’ NLP1342 PHY well suited for next-generation data centers because of the lower latency the device offers, according to the company. The NLP1342 device integrates an enhanced equalizer front end that enables lower jitter, lower latency, and communications over extended cable distances. Optical fiber and twin-ax copper offer lower latency when compared to traditional Cat-5/Cat-6/Cat-7 copper cabling.

In addition to lowering latency, minimizing power consumption is also critical for successful data center deployments given the large amount of high-performance equipment concentrated in a small area. As datacenters move to 10 Gbps and faster speeds, more power is required to transmit at higher speeds. NetLogic Microsystems’ NLP1342 PHY device delivers an ultra-low power profile of 0.5 watt/channel, enabling energy-efficient data centers.

“NetLogic Microsystems’ NLP1342 PHY is the first device for the data center to support 10/40/100 Gigabit Ethernet transmission over twin-ax copper, as well as LR and SR optical cabling,” says Jag Bolaria, senior analyst at The Linley Group. “Twin-ax copper is popular in the data centers owing to its lower latency and lower power consumption than the current 10GBASE-T copper alternative.”

Direct-attach, twin-ax copper cabling offers a cost-effective medium for 10/40/100 Gbps communications over distances of 10 meters. It makes the NLP1342 device well suited for communications from top-of-rack (TOR) switches to rackmount and blade servers. Because the enhanced equalizer front end also supports SR/LR optical cabling, the NLP1342 PHY enables server to end-of-rack (EOR) switch and TOR to EOR switch communications.

The NLP1342 PHY device exceeds the electrical parameters of the nPPI interface as defined in P8023.ba Annex 86A and is compliant with the CR4/CR10 PMDs as defined in P802.3ba Clause 85. The device also offers integration with dedicated pins for SFP+ management and control, dedicated general-purpose I/Os, clock synthesizers, and flexible and programmable Ethernet frame generators and checkers, BER calculators, and link margining tools–all in a low pin-count package ideally suited for SFP+/QSFP+ designs.

Interview with Vladimir Kozlov

Dr. Vladimir Kozlov is the founder and CEO of LightCounting LLC, an optical transceiver and optical interconnects market research firm. Dr. Kozlov has more than 20 years of experience in optoelectronics, optical communications, and market research.

The optical transceiver market is booming—or so says market research firm LightCounting LLC. What’s more, company analysts believe this growth should continue for several more years. Lightwave spoke with Dr. Vladimir Kozlov, founder and CEO of LightCounting, to see what data he had to back up these assertions.

Lightwave: So just how fast is the transceiver market growing?

Kozlov: If you look at the [2010] Q1 revenues of component manufacturers, for publicly traded companies, they were up about 10% on average. That actually shows significant growth, given that Q1 is typically slow.

So it’s growing faster than we anticipated; I think in our last forecast, published earlier this year, we’d been looking at 10% to 15% growth for the year. The way it’s looking now, it’s going to be at least 20% and possibly more for the year. So it’s not the 100% growth that we’d seen in 2000 but it’s probably going to be a healthy 20% or 25% growth for the year, if there is no crash coming—and I don’t think there is a crash coming.

Why do you believe this growth is sustainable?

Well, there are several reasons. If you look at sales of amplifiers and ROADMs, they are really going through the roof and the main suppliers of amplifiers—JDSU and Oclaro, both of them—have been pointing out that the shipment volumes of the amplifiers are exceeding those from the telecom bubble levels. It is a significant data point because it signals the beginning of the next deployment cycle, because the amplifiers and ROADMs are usually installed first onto the system and then the transmitters and receivers are added as network operators light up more and more channels of the network. So if we see the amplifiers and ROADMs market growing over a year or two and growing really fast, it kind of measures the amplitude of the deployment cycle—if the sales of amplifiers are growing for a year or two, that means the sales of other components will probably grow for five or six years just to fill out this initial deployment of amplifiers and ROADMs.

The other factors are kind of more interesting and they’re not obvious. LightCounting has been collecting data on optical transceivers for almost 10 years by now. And the way we’ve been analyzing this data in order to forecast market, we look at what we call an aggregated bandwidth of all the transceivers sold year after year. And if you think about this carefully, this aggregated bandwidth of transceivers—which is basically the units shipped times the data rate of the transceivers—has to correlate with the traffic growth of the network. And our analysis shows that this number for the network bandwidth growth is about 30%—it’s actually a little less, it’s about 25% to 30%. And the data you see on the [actual network] traffic growth is typically 40% or higher. The data that AT&T shares with us shows just under 40%, maybe 38%. Cisco has been monitoring data traffic and I don’t know how they do it, but they’ve been pretty consistent over the last few years saying that, on average, it’s 45%.

So, on the one hand, you have traffic growing at 45% per year, on the other hand, you have network aggregate bandwidth growing at 30% per year. So there is a 15% discrepancy—and the question is, okay, where is it coming from?

So the first issue is, okay, maybe the LightCounting data is not correct. I don’t think that’s the case. So another factor could be that network efficiency is getting better all the time. And I think that’s a valid factor; after all, it would be really sad if all those components are deployed year after year and the network efficiency remains the same. I think the third factor explaining the discrepancy could be continuous under-investment in network infrastructure. And I think that may be one of the main factors explaining the situation.

If you think about it, carriers are certainly in a situation where the traffic is growing, but the revenues are pretty flat because we’re so spoiled; we’re not really paying more for using more bandwidth, or at least not much more. So the carriers are in a situation in which the traffic is growing and the revenues are flat—and, of course, they try to invest as little as possible in the infrastructure to deal with the situation. Obviously, that cannot go on forever, so at some point there has to be a catch-up period where a more intense investment in the infrastructure is required. And that, I think, could be the factor that’s going to sustain the market growth for a few years.

Another factor that is making us more optimistic is the enterprise networking part of the market, where the optics are becoming more critical and the data centers are becoming larger. The data rates are going higher and copper cannot keep up with optics so far at 10 Gbps. So I think the enterprise is going to be another factor sustaining the market for a few years.

Given these factors, which transceiver types should see strong demand in the next few years?

What we’re seeing is that 40 Gbps is doing very well—and it will continue. I think the announcement from Cisco [that it was acquiring CoreOptics] will brighten up the future of 40 Gbps, and what we hear from suppliers is that they’re shipping quite a lot of it.

Tunable XFP modules—I’m talking about the DWDM tunable 10-Gbps XFP—are also very hot products. There still is a little bit of a concern about how they’re going to meet all the specs, but it seems like they’re delivering on the promise [of the technology]. So that could be a good market for a long time.

10G SFP+ for the data center—they’ve been doing very well. Even though 10GBase-T is getting into shape, there is at least a limited window of opportunity—and by limited, I’m saying at least a year or two, maybe three—to capture as much market as optics can and sustain this market. There is a misconception among the end users in the enterprise market that optics is just too fancy and too expensive. Well, there’s some truth to that because some systems vendors make it way too expensive for the end users. But other systems vendors are coming in and delivering it at decent pricing. They’re not marking it up by a factor of 10 anymore; they’re putting a more decent factor of 3 markup, which makes a big difference for the end users.