Enabling 400G Networking with 56G Ethernet PHY IP

Sanjay Gangal interviewed Rita Horner, Senior Technical Marketing Manager of Synopsys at EDACafe headquarteers.

SG: You attended the IP-SOC conference here yesterday, and you did a presentation there.

RH : Yes, we had a very interesting day yesterday at IP-SOC.

SG: Tell us what was your presentation was about.

RH : I presented on the 400G adoption, and the new networking data centers enabled by the 56G PHY. That was the topic of the presentation that I made yesterday.

SG: Tell us about this new hyperscale data centers. What exactly does that mean?

RH : There are many types of data centers, and hyperscale data center are the extremely large data centers. These are the data centers that we refer to as cloud based data centers. These are very large, scalable centers, with thousands and thousands and even millions of virtual machines and thousands of servers in them that are becoming bottlenecked in terms of the throughput because of the high increased demand for the data rate. And all the different needs that we have such as the uplinking of our videos, and different accesses, web searches that we do. It's actually, it's getting slower and slower. A lot of these data centers are implementing accelerators, they're doing deep-learning to be able to process more data, facial recognition and so on and so forth. So they need to have higher and higher throughput. And right now, most of them are adopting 100G data rates and they're getting ready to migrate to 400 gigabit per second data rate interconnect, to be able to process more data through their systems.

SG: And Synopsys had released a 56G Ethernet PHY last year, and is that being used in the new data centers?

RH : That's the enabling technology for the 400Gig. A lot of IC providers and system makers are gearing up the tooling to be able to adopt and start upgrading to 400Gig. So, the 56G PHY that we announced last year is to enable all the IC designers to be able to design ICs that are starting to be used in these new servers, and switches that are targeted for 400Gig. In the coming months, we have customers who are either taping out their designs or starting their next generation 400G products, using 56G PHY in multi-lane aggregation to achieve 400G throughput.

SG: Can you talk about the benefits and characteristics of these?

RH : This technology is a little bit different than the current Non-Return To Zero (NRZ) PHYs and uses multi-level modulation. IEEE adopted PAM-4, which is the 4-level pulse amplitude modulation technique that allows achieving the same data rate using half the bandwidth.

Because the symbol rate with PAM-4 is half of that of NRZ, the signal suffers less from channel loss, and the Nyquist frequency is quarter of the bit rate instead of half of the bit rate that is in NRZ. Standard bodies have already starting working on the next generation 100G ethernet for a single lane.

SG: What are some of the interconnect mediums that support this?

RH : In a data centers, there are connections to the top of the rack or the middle of the rack, where servers to switch connection is typically made with copper cables. Twin-axial copper cables are the most cost-effective interconnects for short connections up to 5meters. And then, rack to rack and row to row connections, active optical cables or multi-mode fibers are typically used for up to 100meters or so. However, for room-to-room, building-to-building, or for connections between data centers, single-mode fiber is being used.

SG: Where do you see this thing going in the next 12 months?

RH : Hopefully, once the optical module providers start ramping the production of their 400G optical modules, we'll see the faster adoption of 400G in the market. Because again, these hyperscale data centers are mega centers with large number of systems. And once they're upgrading from 100G to 400G, they would transition the entire data center with the new interconnect. Therefore, they would need servers, switches, optical modules, updated connectors and cables in large volumes before they could move with their upgrade. If the 400G products are not available in these high volumes, then they would need change their architectures or delay their upgrade plans. Hopefully in the next 12 months, we're going to see the ramp of optical module shipment in high volumes that are enablers of 400G market adoption much faster. Few vendors have been sampling, but we need these products to more to full production before the adoption ramp happens..

SG: What should we expect from Synopsys in the next 12 months?

RH : Our Synopsys 56G PHY for 50G Ethernet market is available today in both 7-nm and 16-nm technologies. Our 56G PHY has a scalable architecture to 112Gig and has all the necessary features to enable single-lane 100G Ethernet. Definitely in the coming months, you're going see more 56G and 112G demonstration of our PHYs in different technology nodes, at different trade shows. And we will be having upcoming technical presentations at different conferences. Please contact us for more information on the foundry and technology node availability as well as hardware demonstration of these PHYs, at Synopsys.com/Ethernet.

Tags: 400G, PHY, Synopsys

This entry was posted on Monday, May 6th, 2019 at 12:49 am. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.