Different sensor chain processing applications have requirements that best aligns them to a particular high-speed fabric protocol. Mercury’s customers are free to choose the best interconnect for their application while maintaining the same processor payload modules. Enabling customers to migrate designs between available switch fabric technologies provides investment protection. It also reduces risk by minimizing the need for expensive application porting and redevelopment efforts.
Mercury offers a wide array of switch fabric solutions that support RapidIO®, InfiniBand, PCIe® and Gigabit Ethernet. Our portfolio of OpenVPX™ building blocks is the broadest and most contemporary within the industry and includes Intel® Xeon® Server-class processing solutions with 40 Gb/s Ethernet and Infiniband switch fabric support.
Mercury’s Intel-based processing modules include POET, which accepts PCIe fabric (the native Intel fabric) and translates it to standard open protocols, including next-generation SRIO and 10 Gigabit Ethernet. Other solutions are Mellanox-based and include support for 40 Gb/s Ethernet and Infiniband. Many of these solutions operate seamlessly with MultiCore Plus, Mercury’s open, real-time embedded system software.
The fastest switch fabric performance across complex OpenVPX subsystems is achieved with Mercury’s Micro Via Radial Interconnect (MVRI) technology. MVRI improves OpenVPX switch fabric interconnect data rates by increasing the signal integrity margin by approximately three-fold, enabling switch fabrics and point-to-point connections to run faster and more reliably.
Mercury’s research has shown that the payload/backplane interconnect is typically the weakest link in an OpenVPX processing subsystem’s signal channel. MVRI technology addresses this bottleneck by ensuring a reliable signaling rate of 14+Gbaud per channel, even across the most complex subsystems. This enables OpenVPX subsystems to run 40Gb/s Ethernet and FDR InfiniBand™ protocols at full speed, providing superior signal integrity. MVRI technology is scalable, enabling it to support signaling rates greater than 14Gbaud per channel.
Signal “Eye” plot without MRVI
Actual signal “Eye” plot with MRVI