Survey says: VPX is the new VME

Our company, which publishes Military Embedded Systems magazine and the venerable VME and Critical Systems magazine, is a voting member of ’s VSO so we’ve been involved in () since it was preceded by (). With the industry’s longest running magazine dedicated to all things VME, we think we’ve got a pretty good handle on the VME market. That was why VITA’s VPX Marketing Alliance came to OpenSystems Media[1] to help facilitate a formal survey between April 23 and July 31 of this year. Digital survey forms were sent to thousands of our subscribers, and a total of 234 people responded, signifying a 65.8% completion rate (154). The results were both predictable and surprising.

The bottom line results

By a huge margin, 57.3% of the respondents are in the Aerospace and (A&D) market, with the next largest segments being Industrial (12%) and Communications (10.3%). While Motorola et al originally conceived VME for all kinds of industries, including European , the bus/board// market primarily has stuck with cost insensitive applications with long life cycles. Today, most nonportable markets choose low-cost boards without backplanes – be they proprietary or open standard. VME’s just too damned expensive. VPX and OpenVPX aren’t much cheaper: Even though higher-density boards replace multiple 6U VME cards, modern gigabit serial interfaces and Core i7 or PowerPC CPUs tend to be a bit spendy.

On the other hand, harsh environments tend to be where VME and VPX play best. This is no surprise, since the IEEE 1101.2 form factor for conduction was adapted to VME’s 6U format by Plessey Systems (which became Radstone) from the format’s inception. VME was needed in systems, and that’s where it’s won favor for decades. Of the 18 questions in the survey, number 12 dealt with Board Formats and Cooling. For 3U sizes, 57.1% of the respondents said that they’d use boards within 12-18 months; the answer was 47.1% for 6U. 

Toss in liquid cooling – a built-in capability for some of VITA 46’s dot specs and accommodated by the Tyco backplane shell with Quick Disconnects (QDs) – and the percentages change to 58.3% for 3U and 51.7% for 6U. By comparison, convection cooling in both sizes was only 22.6% and 29.9%, respectively. What does this mean? VME and VPX, in 3U and 6U, lend themselves well to rugged environments where air cooling isn’t very popular. No surprise, that A&D – and to a lesser extent, Industrial – tends to eschew convection cooling. Blowing a fan across a fire control box mounted in a Bradley in the Iraqi desert isn’t going to achieve much cooling. Similarly, impinging ram air at 40,000 feet trying to cool hot electronics doesn’t work so well in a fighter when the air is so thin.

Technical gotta haves … or not

The early days of VME offered pretty poor I/O density, despite dedicating P2 for user-defined I/O. P0 helped, as did , 2eSST, and , but VPX and OpenVPX are all about moving data around the system. Switched serial fabrics intended for multiprocessor architectures and sensor fusion are the foundation of VPX. Survey respondents ranked high computational density, scalable architecture for expansion, and capable of high-speed backplane signaling all between 65% and 80% of the time. This means that VPX will be used in the high-performance, multiprocessing systems for which it was intended. In fact, 51% of the respondents identified 2-4 processing modules in their system, with a whopping 27.1% identifying 10 or more.

Of the fabrics that were included in Question 13 of the survey, Ethernet, PCI Express, and Serial had the strongest showing. VITA further broke the question down by pipe/lane width (x1, x2, x4, and Not Sure). Ethernet was preferred more than 20% of the time, with x1, x2, and x4 widths fairly evenly represented. PCIe was favored as little as 7% (x1) to as high as 57% (x4). Not surprisingly, Serial RapidIO, still an underdog in VPX, was identified at 7% (x1), 12% (x2), 38% (x4), and with the largest Not sure (43%).

Some of the surprises in the survey results dealt with I/O pin count and the ability to replace modules.  From the get-go, VPX was designed to offer maximum I/O pinout, though the exact number of user-defined varies with module type, size, and OpenVPX profile. A mere 40% of the 98 survey respondents said that I/O pin count greater than 200 contacts was important, but more than 90% said it was a “nice to have.” Still, the question becomes: Was VPX overengineered? Or do serial switched fabrics capable of speeds up to 5 Gbps obviate the need for so many extra I/O pins?

As well, the whole / effort (Mechanical Specification using Ruggedized Enhanced Design Implementation) is meant to be a two-level maintenance version of VPX. That is, sealed LRUs are supposed to be yanked from A&D service by technicians and sent back to the depot or manufacturer for repair or replacement, with new LRUs plugged into service – possibly right on the battlefield. With only 40% of the respondents identifying this capability as important, I’m also wondering if VPX/REDI really isn’t ready. The spec was heavily promoted by Boeing, so it’s possible it will have a limited industry following.

For further study

There’s still lots of confusion in the industry as to the differences between VME, VXS (), VPX, and OpenVPX. At least 40% of the survey respondents gave answers to several choices that said they need more information, more understanding of differences, or more details on suppliers and the overall bus and board ecosystem.

Future technologies identified by the survey include optical backplanes, box-level standardization (such as the three new Small Form Factor proposals being worked right now in VSO), system management (a la ’s ), or signal integrity compliance test definition.

To me, most of the answers in this survey were predictable. VPX and OpenVPX are still very new, with the lion’s share of the activity to date occurring with the vendors working on the standards. Except for the major military primes actively participating in VITA, the word is only now propagating into the marketplace. Clearly VPX is resonating with its targeted user base, as the survey underscores most of what the spec’s creators had envisioned from the start. The Survey has spoken.

Chris A. Ciufo,

cciufo@opensystemsmedia.com

 



[1] To be fair, other industry groups also participated, including my colleagues at M&AE, Journal, Tech Briefs, Extension Media, and others.