Architectures like the VMEbus have been so successful largely due to their continued performance enhancements while maintaining backward compatibility. Over the years, the steps have been incremental for these types of backward-compatible open-standard (BCOS) technologies. But today, new innovations can take VME and CompactPCI to revolutionary levels.
BCOS technologies have let users employ existing cards while migrating to higher bandwidths. Users also get the benefit of existing software, drivers, and knowledge-of-use, which can be a large portion of a technology’s return on investment.
Take VME for example. VME evolved from 16-bit J1 at 20 Mbytes/s of bandwidth in the early 1980s to 64-bit VME64x at 160 Mbytes/s in 1997. In 2001, CompactPCI combined backward compatibility and improved performance with Gigabit Ethernet (by adding switched-fabric routing across the backplanes) in the PICMG 2.16 specification.
The VMEbus International Trade Association (VITA) followed suit in 2003 with Gigabit Ethernet over VME64x (VITA 31.1). This took an otherwise VME64x backplane, added a switch card slot(s), and used previously undefined signals in the backplane’s P0 connector to run the fabric in star and dual-star topologies.
VXS followed later in 2003, with the same concept as VITA 31.1, except that it replaced the 2-mm HM P0 connector (which experiences performance declines above 1 Gbit/s) with the MultiGig RT-2 connector (which performs well above 6.25 Gbits/s). VXS also added features for XMC modules, conduction cooling options, system management, and more. Standard centralized VXS topologies offer approximately 20 Gbits/s of aggregate bandwidth.
On the VITA side, VITA 46 (VPX) has been introduced for high-performance applications. It adds a mesh configuration where each slot acts as a hub slot for very high bandwidth. It uses multi-gigabit connectors for all of the slots, generally forgoing backward compatibility.
VPX also was specifically designed to go with a new enclosure initiative, VITA 48 (REDI), which offers advanced cooling options and a very rugged design. The architecture likely will have a role in many 3U (and 6U) designs where high bandwidth is required within less space for rugged applications, such as unmanned aerial vehicles. It’s possible to incorporate a degree of backward compatibility in the backplane.
On the PICMG side, AdvancedTCA was introduced to serve primarily the telecom central office market. Like VPX, it was designed with a new high-speed connector. ATCA uses the ZD connector for 6.25-Gbit/s and beyond performance. The form factor also increased to 8U by 280 mm for more space on the processing blade. Redundancy and shelf management were built in to ATCA to serve the high-availability market. Aggregate bandwidth is up to 1.2 Tbits/s.
While these technologies are important, don’t forget the benefits of backward compatibility. A mesh configuration over VXS called VXS Processor Mesh recently has been developed for VME. It defines the pinouts for a 4x mesh configuration for up to 112 Gbits/s of bandwidth, which is nearly a sixfold performance increase over typical VXS. However, it still retains compatibility. Also, the migration to switched fabrics like PICMG 2.16 extended CompactPCI’s life. Now, PICMG is following VITA’s VXS example by replacing some of the legacy connectors with high-performance connectors. CompactPCI Express adds the ZD connector for higher speeds, but it also can retain legacy slots for backward compatibility. The aggregate bandwidth is up to 40 Gbits/s.
With new technologies like VXS Processor Mesh and CompactPCI Express, VME and CompactPCI have retained the important feature of backward compatibility while offering a massive increase in performance.