In this episode of the Futurum Tech Webcast – Interview Series, we focus on the topic of coherent digital coherent processor (DSP) innovation and why it is vital to enabling cloud providers and carriers to replace traditional transport equipment for improved scalability and total cost of ownership (TCO). Marvell’s new Orion coherent DSP offering provides twice the bandwidth over current DSPs for pluggable modules, Orion can increase the performance and reach of pluggable module-based connections up to ~2,000 km. Plus, Orion broadens the opportunity to deploy coherent DSP-based modules in more use cases, including 800 Gbps metro networks, 600 Gbps regional networks, 400 Gbps long-haul networks and 800 Gbps data center interconnects (DCI).
My guest today is Samuel Liu, Senior Director, Product Line Marketing at Marvell Technology, a top-tier semiconductor company. Samuel is a new guest on this show, and astutely shares his insights and perspective on the direction of the optical transport market segment and why Marvell’s coherent DSP portfolio is driving new breakthroughs.
To start our discussion, we focused on the debut of the Orion product and why it’s the industry’s first 800 Gbps coherent DSP for pluggable modules. Samuel spotlighted how pluggable modules can lower capital costs by up to 75% compared to traditional equipment as well as significantly decrease power and space. The 800 Gbps threshold represents a turning point in the industry by doubling the bandwidth and is applicable to over 80 percent of use cases throughout cloud and service provider environments. Plus, Marvell’s new solution already has broad ecosystem support from module players such as Fujitsu, Lumentum, Coherent, Innolight, and others as well as Marvell offering its own COLORZ 800 module.
Our conversation examined the following key topics:
- How the Marvell’s Orion coherent DSP solution supports DCI over long distances including up to 2,000 kilometers.
- The key distinctions between the cloud and service provider use cases such as bolstering metro, regional, and long-haul capabilities for service providers and strengthening DCI capabilities in areas such as more availability zone flexibility for cloud providers.
- Why pluggables were developed and how ZR/ZR+ innovation is helping to drive pluggable module innovation.
- How service providers are driving demand through rapid upgrades to 400G+ routers as well as how data centers are fueling demand through upgrades to 51.2T capabilities.
- Why AI is playing a key role due to massive demand for AI training in centralized locations and AI inference across distributed, regional data centers requiring DCI advances to ensure application optimization. Plus, the emerging role of virtual hyperscalers.
- A look into what’s next for Marvell’s Orion portfolio and coherent DSP technology including the expectation of greater experimentation by service providers and cloud providers.
You can watch the video of our conversation here (and subscribe to our YouTube channel while you’re there):
Or listen to the full audio here:
Disclaimer: The Futurum Tech Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.
Transcript:
Ron Westfall: All right. Good day everyone and welcome back to the Futurum Tech podcast. I’m Ron Westfall, your host, research director here at the Futurum Group. And I’m excited for today’s episode because we’ll be talking with Samuel Liu, and Samuel comes from Marvell Technology, and today we’ll be diving into Marvell’s innovative breakthroughs in the coherent DSP technology area. So, with that, Samuel, please introduce yourself, you’re a first timer, and I know we’re going to have an exciting conversation ahead. And, so with that, tell us about yourself and what brought you into the industry?
Samuel Liu: Thank you, Ron. I’m very glad I have this opportunity to join the discussion. My name’s Samuel, Samuel Liu. I’ve been working in the network equipment industry for more than 25 years. Most of that time, I actually worked for this big name router, transport company, and some startup company. So for the last five years I’m honored to have this opportunity to manage coherent DSP as a PLM in Inphi and then in Marvell. So all from the 16 nanometer, seven nanometer, now we announced this new product, an exciting product, a five nanometer product.
Ron Westfall: Yeah, thanks for that Samuel. That’s very helpful background because we know that Marvell has come out with a couple of new important announcements, specifically the announcement related to Orion 800 and also the COLORZ 800 module announcement. And I know the folks in the industry out there want to know more about what is going on here, in terms of coherent DSP technology innovation, and what Marvell is bringing to the industry. And, with that Samuel, what is it that is really coming out that is going to be important for folks to know?
Samuel Liu: We announced two major products in the last few days, last week. So the Orion is the new generation coherent DSP, based on five nanometer technology, which has hit the market. It’s the first to enable coherent pluggable modules, and we can build a solution to address both carrier use cases and the major hyperscale cloud data center in the connection solutions. At the same time, Marvell also announced our internal pluggable module based on Orion. So that’s really laser focused for enabling the high speed 800 gig ZR/ZR+ cloud pluggable module to address the cloud use case.
Ron Westfall: And yeah, that certainly I think is important for folks to know. And what also leapt out at me were the new distance capabilities. For example, being able to go anywhere from two kilometers all the way up to 2000 kilometers, if you will, the London to Minsk connection that comes to mind. And, with that, can you elaborate more, Samuel, on what are these distance capabilities that are built into the new products?
Samuel Liu: Yeah, so distance it’s important, it’s easy to remember, right? So on the carrier side, the use case for example, based on the CFP2-DCO pluggable solution enabled by Orion, we can cover all the way with a 400 gig, all the way to 2000 kilometer, as we call the typical, long-haul use cases. But we also offer many different variations of solutions to optimize for regional, give or take around a thousand kilometers, or metro, right? Both 400 gig and 800 gig. On the other side, carriers also start to adopt this new architecture called IP over DWDM. We collect the layer, for the packet layer and the transport layer, to allow the small form factor pluggable solution, directly plug in on the router, in the field. With that one, we also offer some new solutions. So we work closely with OpenZR+ to define those 400 gig IP over DWDM solutions, beyond the metro. We are talking about a thousand kilometer regional and also 1500 kilometer and beyond for long-haul. So those are the distances covered with the higher data read, 400, 600, 800 gig, use the small form factor pluggable solution.
Ron Westfall: Now that’s very exciting. I’m glad, Samuel, that you mentioned pluggable optics because that’s clearly a technology that is garnering a lot more attention. And, with that, what is it that’s driving the decision making with, on the one hand the cloud service providers, and on the other the carriers? What is it about pluggable optics that’s generating so much interest across the industry?
Samuel Liu: Yeah, so the pluggable solution itself has a clear benefit. So it’s like they can allow the end user, we call the Pay As You Grow, so you can reduce the initial investment, gradually go and add it on. You can make the network more reliable, simpler. So if one of them broke, or two, all right, so you can easily just deal with that port, instead of traditionally, if you stretch the density with one fixed line hard with multiple ports, one of them broke, you might try to swap the whole line hard.
So there are many, many benefits, but also on the other side there are many challenges because you need to offer the good performance and the small enough power to fit in for this small form factor pluggable. So that’s one major aspect. On the other side, pluggable sometimes also coupled together with what we call the interoperation. Because of pluggable solutions, we can make it like our own Marvell Technology, with the cutting edge, with the best performance, and the best power for feed-in pluggable. But many end users also start to ask, “Since I have the pluggable solution, I’m driving what we call the disaggregated architecture.” Right?
Ron Westfall: Right.
Samuel Liu: They’re not necessarily bundling one solution with one OEM. They may just open to see, hey, they can deal with module vendors A, B, C, they all offer a similar solution. You plug on this side router, on the other side, they can talk to each other. It’s an interoperation of pluggability. But on the other side, the cloud, as you mentioned, the cloud DCI or the hyperscale, even from the current generation, the 400 gig generation, they already start to majorly deploy with this new transport architecture. That’s the IP over DWDM, right? So you kind of combine the two major network equipment nodes, the router switch and optical transport, more like into one. So you directly plug in the small form factor, pluggable module on the router switch. So you majorly reduce the investment of CapEx, the OpEx. You reduce the space, you reduce the power, you make the transport lead much simpler, much more reliable. So these few things all interrelated.
Ron Westfall: Mm-hmm.
Samuel Liu: But, from a different angle, to give these comments. So…
Ron Westfall: Yeah, I know the pluggable optics and the pluggable modules are directly linked to those emerging use cases that you touched on, Samuel, and I know I see exciting ones on the horizon. And so let’s drill down, what about the use cases that are being driven by these new Marvell products?
Samuel Liu: So let’s start with the cloud use case or the cloud DCI use case. So for them, when they grow their bandwidth, grow the switch capacity from the 400 gig per port with the current 12.8T switch, either they double it or many of them are even tripling or directly quadrupling to 51.2T switch. With the big capacity, they still need, what we call, the data center interconnection. The physical size, physical distance is still there, the fiber, the patch panel is still in the field, but now they need to connect them together instead for 400 gig, they need to use 800 gig. So this is the one clear use case to support the cloud player, their capacity upgrade.
The other one, on the carrier side, their upgrade cycle is longer. So it’s more like every once a decade, every eight, 10 years. But they are going through, starting from this year, a major upgrade from 100 gig router to 400 gig router. The core router, the edge router, service, volume and aggregation. So when they upgrade the router, they want to have the data read end to end, right? So no matter if it is a 2000 kilometer, thousand kilometer or 500 kilometer, if it’s a 400 gig device, we keep it in 400 gig. So there’s a huge benefit from the packet layer due to the load balancing and other considerations. So with this and the compound together with the CFP2 pluggable module, finally by Orion-enabled, they offered performance. It’s the same or even better than the bulky line hard-based, not pluggable obviously, right? It’s a lower cost, lower power, all these benefits, even with the current optical transport network architecture.
But on the other side, some carriers are going to start to think, “Hey, why hyperscale cloud DCI? They adopt IP over DWDM and they build a lower cost, more reliable, simpler network. Why not, we also do it?” So some of the carriers started to, more likely, instead of evolving to use the current traditional architecture, they started to adopt this new IP over DWDM architecture. So that one requires beyond just 400 gig ZR, also called ZR+. They need a longer distance. So they need higher output of power for the module. It’s not just regular OIF defined, negative 10 dBm. So they need an integrated amplifier to directly come out, zero dBm, give or take. They can fit in for the existing optical transport gear, the Rodin in the field. And the longer distance for them is not only increasing the engineering margin for the deployment, they’re also going to cover more and more use cases where you have better performance. So from both angles, Orion enables those new solutions to be available in the next few months, for the carrier to consider.
Ron Westfall: Yeah, and with these new Orion 800 capabilities, I think this is very exciting and this can now be an opportunity to talk about, how did pluggables, themselves, come about? And Samuel, I know you have some definite thoughts on this and about what’s driving the module movement, and for that matter, now, what about a quick history of ZR/ZR+ technology? What is the background for that as well?
Samuel Liu: Yeah, if we rewind it back, the industry, five years back, right? So an OIF started to define this new category called the ZR WDM 400 gig. And of course this comes with a lot of fighting, but now you can see the whole cloud DCI architecture, it’s become a default architecture. All the major ones, they introduced the use of 400 gig ZR, some 400 gig ZR+ plus, directly plug on switch of 12.8T switch. So this is where the ZR actually came from and hit the market with a huge success, to a large extent thanks to not only my coherent DSP, also my internal module product they offer. So to help shift the whole industry to move to this way, they can see this is totally tangible, the benefit, because architecture-wise, this is the ZR WDM.
As I mentioned, you’ll simplify the transport and you’ll save money from CapEx, OpEx. But on the other side, the hyperscale DCI, they need a huge volume. So this is a lot of inner-workings, not only the technology itself, even the manufacturing, the yield, there’s a lot of things that need to all work together. So ZR for cloud DCI, it’s become a default. Now, the ZR concept or ZR+ concept starts to move to the carrier side.
They also want to take advantage of it, but they have few immediate questions like, okay, my use case, most of the time it’s beyond just ZR, per say, the 80 kilometer, hundred kilometer. They need four, 500 kilometers, they need a thousand… They need beyond. That’s a ZR+ cycle down, which if a metro distance is around four, 500 kilometers. But there’s no buzzword for that, okay, so what do you call a thousand kilometers, right? So is it regional, back to the traditional terminology for long-haul, regional, metro? So that’s the way to consider. And on the other side, I think they also need beyond the OIF to define 400 gig ZR, 800 gig ZR. So we work very closely, actually, drive the OpenZR+, and I say, we define not only the longer distance, but also the higher output of power. So allow this small form factor coherent pluggable module 400 gig, 800 gig, to cover more and more use cases.
Ron Westfall: Yeah, with those capabilities, I see the demand is really going up. It’s become, I would say, somewhat urgent. And I know in my discussions, with the carriers and the cloud service providers, that there’s, quite simply, more interest in expanding adoption of coherent optics and, quite simply, taking advantage of the capabilities you outlined there, Samuel. And so from your view, how urgent is the demand? What is driving the adoption of the new Marvell products?
Samuel Liu: Is very urgent, overall, but again we need to separate to talk about the cloud market segment versus carrier market. The cloud market segment, they are going through this new cycle of the switch capacity upgrade, as we mentioned from 12.8T switch of 32 by 400 gig, they can double it and they can quadruple it directly to the 51.2T switch. So with that upgrade, they still need the data center interconnection to match their switch capacity. 800 gig ZR/ZR+ is the only solution available to support, and do they need 1.6T? Yes, they need but the whole industry needs more time for technical innovation, the technology evolution and then putting things together a few years down the road. For this generation in the next few years, 800 gig ZR/ZR+ is the solution to enable the hyperscale upgrade for the switch.
On the carrier side, it’s different, right? They’re upgrading for the 400 gig route, it’s even faster. It’s ongoing faster, even though it’s every eight, 10 years upgrade. But their 400 gig router deployment for the Tier 1 carrier already started this year now. So it will take a few more years, then the Tier 2 starts the next year. So this whole thing, when they upgrade the 400 gig router, it’ll stay there for five, eight, 10 years. Long time. So to some extent, the carrier use case is more immediate, either it’s the traditional long-haul pluggable CFP2 or the IP over DWDM we talk about, right? So with carriers– both ends, they have a strong demand and to some extent they’re waiting for an optical solution, waiting for the coherent DSP-enabled pluggable solution.
Ron Westfall: Yeah, no doubt that both the carriers and the cloud service providers are keen on how to better handle and scale their AI workloads. And that’s been the hot topic of the summer and I anticipate it’ll be an ongoing hot topic for the rest of the decade. And as such, it’ll certainly be very interesting to know, what about Marvell’s expanding role in enabling the support of AI workloads? What do you see out there, Samuel, that folks are emphasizing and how Marvell can make a key difference in this particular area?
Samuel Liu: Yes, AI is exploding for the last few quarters, but most people understand the AI within the data center, with a shorter reach distance. And on that front, obviously, Marvell’s right on the line to enable from a few meters to a few hundred meters, a few kilometers. Within the data center we have all these different solutions. It’s a PAM4-based DSP. But on the other side, this AI-driven growth within the data center, will not stop only inside the data center. It will eventually, or it’s ongoing, to drive between data centers. You need the data center interconnection to grow with the bandwidth. So it’s like you can call it distributed because one data center is not enough to finish all the computation, and that’s from the AI, machine learning, the training itself.
But on the other side, as a user, you also want to plug in, want to take advantage of AI capability and more users start to also. So users are all distributed with the distance, so keep driving what we call the data center interconnection, from a longer distance along ZR+, to ZR, or nowadays start to pop out a new category ER, or those shorter distances, but it’s outside of one data center. The bandwidth starts to grow and that’s the part, Marvell, we start to offer with Orion, the coherent DSP, also our own pluggable module, 800 gig ZR/ZR+.
Ron Westfall: Yes, and as a result, I know folks out there are going to be keen about Marvell’s roadmap. What is next? What can we anticipate, Samuel, in terms of what Marvell is planning, in terms of enhancing and improving the portfolio capabilities?
Samuel Liu: Yeah, again, so there’s two market segments, right?
Ron Westfall: Mm-hmm.
Samuel Liu: So on the cloud side, their bandwidth growth, all the network equipment upgrade cycle, it’s like a three, four, five year in that range. So in the next few years, the 51.2Ts become a new norm and people start to even talk about it beyond a hundred T. So the next generation ZR/ZR+ interconnection, like a 1.6T, will be demanded. We are definitely focusing on, to be the first down the road, to go along together with the whole industry, to offer those solutions, to support those bandwidth growth, their upgrade. So that’s one direction we’re considering and we’re fully committed to driving. And the other ones, on the carrier side, we need to think about what is the best way going forward.
Ron Westfall: Right.
Samuel Liu: As I mentioned this router upgrade maybe happens once every decade, every eight, 10 years, right?
Ron Westfall: Mm-hmm. Right.
Samuel Liu: So, 400 gigs, is probably going to stay for a long time in the field, and then what will be the next, right? We can continue building solutions to extend even more for performance, at an even further distance. We can drive a lower cost, lower power, so we can actually come up with some optimized design, even when they push the 400 gig from metro, from aggregation, even go down to backhaul eventually. So there’s a lot of different things we need to consider. But on the other side, 800 gig routers start to come in, for some big, large core node. So those will start to be available in the next few years. 800 gig performance, how do we enhance, like cover longer distance, better performance, still fit in for pluggable modules?
Ron Westfall: And, yeah, I know I’m looking forward to what is being developed and certainly I see it aligning with what again, the cloud service providers, the hyperscalers, are demanding and prioritizing, as well as naturally, the operators, in terms of what they are actually going to spend more money on eventually, in terms of upgrading all of their router capabilities, certainly when it comes to leveraging pluggable optics. And I know that is going to be a huge difference maker over the next couple of years, plus. And, with that high note, thank you so much, Samuel, for joining our webcast today.
Samuel Liu: Thank you Ron. Very glad to have this opportunity to discuss our Marvell solution.
Ron Westfall: Exactly, and with that, everyone please hit the subscribe button, join us for our upcoming Futurum Tech podcast. And this was a great conversation. I know that Samuel definitely brought in the insights that are going to be very valued by the audience out there, particularly when it comes to Marvell’s contributions to coherent DSP technology. And again, we appreciate Marvell for joining our webcast today, and with that, good day everyone and thanks for tuning in. We’ll see you later.
Other insights from Futurum Research:
Marvell Q2 Fiscal 2024: AI and Cloud Are Top Growth Drivers
Marvell Launches COLORZ 800: ZR/ZR+ for Spanning the Cloud
Marvell Powers Automotive Ethernet Switch Technology into Global Prominence | Futurum Tech Webcast
Author Information
Ron is an experienced, customer-focused research expert and analyst, with over 20 years of experience in the digital and IT transformation markets, working with businesses to drive consistent revenue and sales growth.
He is a recognized authority at tracking the evolution of and identifying the key disruptive trends within the service enablement ecosystem, including a wide range of topics across software and services, infrastructure, 5G communications, Internet of Things (IoT), Artificial Intelligence (AI), analytics, security, cloud computing, revenue management, and regulatory issues.
Prior to his work with The Futurum Group, Ron worked with GlobalData Technology creating syndicated and custom research across a wide variety of technical fields. His work with Current Analysis focused on the broadband and service provider infrastructure markets.
Ron holds a Master of Arts in Public Policy from University of Nevada — Las Vegas and a Bachelor of Arts in political science/government from William and Mary.