Q&A with Mercury Systems about the Intel Design Solutions Network
Making the most of Intel DSN strengths
Editor’s Note: Earlier this year Altera, now Intel® Programmable Solutions Group (Intel PSG) launched the Design Solutions Network, or DSN. According to the company, DSN aims to help customers bring innovative solutions to market more rapidly. DSN connects customers with DSN members—including defense electronics solutions provider Mercury Systems— a Platinum-level member offering products or design services using Intel PSG CPLD, FPGA, SoC and Enpirion power devices.
Recently I had the chance to ask Rich Jaenicke, Senior Director, Strategic Technology and Alliances, Mercury Systems, to put the Design Solutions Network launch in context for Embedded Intel Solutions readers. “Our Platinum-tier membership in DSN brings a higher level of engagement and support,” Jaenicke says. He adds, “and Intel as a supplier moves closer to being an extension of our engineering team. The combination of Intel CPU and FPGA technologies is now even more advantageous for our customers, as we can now offer them better integration and support.” Following are excerpts from our discussion:
Embedded Intel Solutions: What FPGA related topics deserve more attention from your perspective?
Rich Jaenicke, Mercury Systems: Two areas that have some traction but could use more attention are FPGA offload for networking and for security. FPGA-based offload for networking is becoming more important because the line rates and protocols are changing more rapidly. Do you upgrade from 10G Ethernet to 40G or go directly to 100G? How many upgrades do you really want to do?
By having an FPGA-based NIC [Network Interface Card], you can upgrade in firmware as often as required without needing to invest in new hardware for every endpoint. Some leading-edge data centers take this approach, and it is readily applicable to Mercury’s targeted defense market as well.
Similarly, there is an increased emphasis on the level of security and the reaction speed to new threats and discovered vulnerabilities. Software-based security often does not have the level of performance needed for modern threats without consuming the majority of CPU cycles. FPGAs are faster than CPUs for these types of tasks, and updates to new threats or vulnerabilities can happen in days. ASIC-based approaches could take years. It also helps if there is native support for a variety of security building blocks in the FPGA, and the Intel PSG roadmap has a compelling story there.
Embedded Intel Solutions: The news of Mercury’s selection as a Platinum-tier Provider in the Intel DSN mentioned “pre-integration of server-class Intel® Xeon® E5 processors with the latest Intel PSG FPGAs into SWaP-efficient, rugged, scalable modules.” Can you elaborate on that?
Rich Jaenicke, Mercury Systems: Yes, Mercury Systems has been integrating server-class Xeon CPUs (Figure 1) and Altera® FPGAs since 2010, when we combined an Altera Stratix® 4 with two quad-core Xeon 5000 series CPUs. Today the FPGA acts as a protocol-configurable fabric bridge and a protocol offload engine for the Xeon CPUs. That approach of combining FPGA and Xeon technology foreshadowed the multi-chip package (MCP) Intel disclosed earlier this year that combines a Broadwell EP Xeon processor and an Arria® 10 GX FPGA.
Mercury invested in advanced packaging and cooling technologies to deploy the combined FPGA and dual Xeon solution in a rugged, size-constrained, Open System Architecture module. Breakthrough packaging technologies include repackaging Xeon E5 LGA processors as BGAs and developing direct-attach DIMMS. Those capabilities enable rugged, dense packaging and efficient cooling with full-speed processing.
Embedded Intel Solutions: What is the best way to leverage the advantage of early device access?
Rich Jaenicke, Mercury Systems: Early access to the latest exciting technology is sometimes intoxicating, which can lead to over-optimizing around the idiosyncrasies of the early access device and its toolset. If you spend extended time finding workarounds for all the early bugs, you will waste resources and may even delay the release of product using the production device. A better approach is to run a separate program for the early access device solely to provide a learning experience that will reduce risk and accelerate deployment of the production device.
Embedded Intel Solutions: How does Intel philosophy and track record factor into success, whether for markets that use FPGAs or anything else?
Rich Jaenicke, Mercury Systems: Intel manufacturing technology and philosophy have a significant potential impact on future FPGA products. Access to Intel’s US-based fabs has significant advantages for security and a trusted supply chain. Access to Intel’s manufacturing technology for fabrication, packaging, and testing should have positive impacts on quality, performance and time-to-market for particular FPGA product lines.
Embedded Intel Solutions: Closing thoughts?
Rich Jaenicke, Mercury Systems: We hope that some Altera philosophies make their way into Intel as well as vice versa. Altera had a stronger focus beyond the data center that has led to a broad selection of devices with extended temperature and longevity of supply. Intel’s Internet of Things Group (IOTG) does that to some extent, but we would like to see Intel expand the offerings more deeply into the higher-end processors, with those requirements becoming part of the design. Already delivering advantages to defense and industrial applications, such products could also enable the data center to require less cooling and even become mobile.