EXTRA! OpenVPX goes from conflict to collaboration
It’s all over folks: OpenVPX becomes VITA 65; vendors announce boards and systems. The next generation of the venerable VMEbus is born.
Just about the most exciting thing that’s happened in the VME ecosystem this year has been the controversy surrounding OpenVPX, VITA, and myriad companies with skin in the game. Recall that a gang of rebels egged on by Mercury Computer Systems started the OpenVPX Working Group back in January because they were appalled by the lack of interoperability built into the 3-years-young VPX/VITA 46 specifications (OpenVPX Industry Working Group: Open for business, or just controversy?). “Shockingly,” Mercury and friends wanted to follow the PICMG model of control-data-management planes to assure high-availability military systems that would accommodate LRUs from multiple vendors with minimal backplane changes.
The way VITA 46 evolved – with Curtiss-Wright Controls Embedded Computing (CWCEC), Mercury, GE Fanuc, Tek Microsystems, and many others driving the specs – allowed for too much vendor-specific pinout customization. Mercury (and their customers such as Boeing) essentially argued that once a VITA 46 design was won, it was damn near impossible to design out the winning vendor from a quasi-sole-source position (Mercury's 'OpenVPX Industry Working Group' colors outside VSO's lines: An efficient technology fast-track or Pandora's box? Q&A with Mercury Computer Systems' Greg Tiedemann). At the time, I applauded Mercury’s stake in the ground because I’d previously railed against this “feature” in the VITA 46 specs.
But besides breaking ranks with the VITA 46 team in the VITA Standards Organization (VSO), what made OpenVPX such an evolving soap opera was that VITA’s executive director Ray Alderman openly endorsed the OpenVPX Working Group’s efforts, which seemed totally at odds with Alderman’s own organization. (Those VPX surprises keep a’coming, and Alderman’s own OpenVPX: It’s all about the backplane). Alderman also seemed to tacitly endorse PICMG’s architecture as the basis for OpenVPX. Oh, the horror! And if all this weren’t enough (think: “splinter group breaks ranks and relies on VITA leader’s support”), the OpenVPX team refused to let CWCEC join the party and play in their sandbox. Since CWCEC appeared to have more VITA 46/VPX design wins than anyone, their customer base fired back at the OpenVPX Working Group with hate mail aplenty (VITA 65 versus OpenVPX: Why can't we all just get along?).
Under extreme pressure (and after signing a few secret NDAs and agreements), CWCEC was admitted to the OpenVPX Working Group and even agreed to chair the VITA 65 committee that would eventually re-integrate OpenVPX spec results into VITA. (Are you still with me? I said it was a soap opera.)
In the end – the entire VME ecosystem was working together outside of VITA to realize Mercury’s vision of interoperable VPX boards – now just called “OpenVPX.” By Fall 2009, the plan was to present the specs back to the VITA 65 committee and dissolve the OpenVPX Working Group (The best way to run a railroad). It was a flawless example of open standards, competitive collaboration, and how VITA continues to be relevant while so many other standards organizations struggle to make a market impact.
The present: MILCOM announcements
I’m happy to report to you that everything went as planned, and yesterday (October 19, 2009) at the AFCEA/IEEE MILCOM show in Boston, MA, the OpenVPX Working Group presented to VITA the results of nine months of public collaboration resulting in a set of interoperable OpenVPX specifications with huge vendor and customer support. The VITA 65 committee will now formalize for ANSI approval a set of documents that builds upon the previous VITA 46 (VPX) and VITA 48 (REDI) specifications, but adds air-cooled LRU plug-and-play vendor interoperability via Rear Transition Modules (RTMs). The specs define a complex set of profiles to assure that one vendor’s board can be replaced by another vendor’s board with only a minimal backplane change.
As you might expect, the key movers and shakers in this little drama also announced plans and products at MILCOM.
Like proud new parents with their first child, CWCEC and Hybricon announced (last Sunday night – before MILCOM kicked off) what they claim is the first “live OpenVPX system.” Yesterday, Mercury – the instigator of this whole tale – announced a series of new “V1.0 OpenVPX Specification” products.
CWCEC and Hybricon’s OpenVPX mission computer
Proving VITA 65’s vendor interoperability will require modules from one vendor to be swapped out and replaced by another’s. But to start, CWCEC and Hybricon demonstrated at MILCOM that real working VITA 65/OpenVPX systems are available and ready to solve military problems. Comprised of a four-slot vetronics-style cold-plate chassis, the Hybricon SFF-4 Small Form Factor box swallows 3U OpenVPX-style cards plus a plug-in PSU (Figure 1). Of interest is the lack of any cabling in the box; Hybricon relies on a plug-in front panel PWB for mounting the 38999 connectors.
Based upon the VITA 48 REDI 1-inch pitch backplane, the chassis is designed for MIL-STD-704F aircraft and MIL-STD-1275B vehicles. In the MILCOM demo configuration, it’s populated with two 3U OpenVPX PowerPC SBCs (VPX3-127), an FPGA module (VPX3-450), a carrier card for PMCs or XMCs, and two mezzanine modules: one with an eight-port GbE switch (PMC-650) and the other with a dual output graphics controller (XMC-710). All modules are rugged and conduction cooled (Figure 2). Collectively, the cards talk on several OpenVPX backplane profiles that incorporate data, control, and management planes based upon Mercury’s original vision of OpenVPX and implemented as payload, peripheral, and switch architectures.
The live demo showcases production-ready products. Hybricon’s chassis, originally scheduled for production in mid-2010, was accelerated by more than six months to meet market demand. And CWCEC’s modules, though redesigned from the previous VITA 46 “VPX” to the new VITA 65 “OpenVPX” flavors, are representative of all the company’s modules transitioning to OpenVPX. That is: CWCEC says all of their cards are easily migrated to OpenVPX from the previous VPX versions. What’s most interesting to me is not that this demo is a production-ready mission computer or HMI display controller – it’s how quickly CWCEC went from being the underdog at the OpenVPX Working Group table to being the first company to announce working OpenVPX modules. (CWCEC says 6U-sized cards will be available soon.)
They beat Mercury by a scant 24 hours.
Mercury Computer Systems at MILCOM: “Broad range of products”
You’d expect Mercury Computer Systems to announce a bunch of OpenVPX cards at MILCOM, since it was their idea in the first place. Compliant with the “new V1.0 OpenVPX and draft VITA 65 Specifications,” Mercury’s Ensemble 3000 Series 3UOpenVPX and Ensemble 600 Series 6UOpenVPX product lines comprise SBCs, switch and I/O modules, and high-compute density boards – the kind for which Mercury is best known.
While no specific cards were announced at MILCOM, a trip to the company’s website lists six 3U OpenVPX choices, from complete systems to Virtex-5 FPGA nodes and OpenVPX Ethernet switch hubs, to RapidIO data plane switches. In the 6U category, Mercury has nearly 10 products listed, from 8640D dual-core PowerPC SBCs to serial FPDP I/O modules.
What’s it all mean?
Clearly VME is migrating to the serial fabric I/O “goodness” of OpenVPX. But technology isn’t enough in systems where customers want multi-sourced choices. The OpenVPX Working Group, founded by Mercury Computer Systems and driven to success by heavyweight Curtiss-Wright Controls Embedded Computing and their partners such as Hybricon, has shown that breaking the rules can accomplish great things quickly. By moving outside of VITA, the 28 OpenVPX Working Group members have created an open standard that will soon become an ANSI standard under VITA 65.
The controversy’s over. The conflict is resolved. Collaboration among competitors has shown that VME’s future is in OpenVPX under VITA 65.