VME: Does it Have a Future?



There are several reasons the answer is a resounding “yes.”

In the defense and aerospace market, VME has enjoyed widespread adoption over the years. But: will it continue to be deployed? Is there still life in the old dog? Will manufacturers like Abaco Systems still be manufacturing VME boards in 20 years’ time?

The answer to those questions is a resounding “yes.”

The Importance of VME
So why is VME still so important in the defense and aerospace market? There are many factors, but at the core is the need to ‘deliver’ on programs with as few perturbations and surprises as possible along the way. The name of the game has always been to minimize, mitigate—eliminate, even—risk.

VME ticks those boxes. It isn’t, of course, the only technology that can be delivered with this level of minimal risk, but it is widely regarded as an absolutely safe choice—and that counts.

Figure 1: The M1 Abrams Tank is an example of a program with 30+ years of service life.

The Rise of VPX
VPX is a related standard and one of the many that have been seen as the ‘replacement’ for VME, in the same way as CompactPCI and AdvancedTCA were trumpeted in the past. The difference with VPX is that it has actually taken substantial roots in the defense and aerospace market and is an entirely viable alternative to VME. It has at least three significant selling points—

  1. It offers a route to higher performance through its fabric-based interconnect
    VME is built around a parallel-based bus, which is limited in speed and can be a bottleneck in applications that require large data movements between boards. VPX removes this bottleneck as it is a point-to-point, fabric-based architecture, and can therefore offer more flexibility and performance in its interconnect than VME.
  2. It has a viable small form factor (3U) which is well-suited to size-constrained platforms
    VME in its 3U form factor (160mm x 100mm) has no rear I/O, so it has extremely low applicability in most rugged defense and aerospace applications. VPX fixed this problem by providing rear I/O pins and is now very popular for applications such as UAVs where space is extremely limited.
  3. It has many of the same physical attributes as VME
    By keeping the same definitions of board sizes, with 3U and 6U (160mm x 233mm) options, VPX is well positioned to slot in the same spaces as either 6U VME or 3U CompactPCI—both of which have been popular in defense and aerospace applications

So why doesn’t everyone just choose VPX rather than VME? Again, there are many things to consider.

Performance
Although higher performance seems like a good idea, not all applications require high bandwidth communication between boards. For example, an application such as a missile warning system has to do “a lot of thinking about a small amount of data,” as one of Abaco’s customers once characterized it, so adding fabric-based interconnect is effectively only adding to the power consumption of the system. High speed interconnect, whether embedded on the CPU or an external switch, tends to be power-intensive—whereas VME is not.

Risk
VPX is no more risky than VME from a vendor’s point of view. Abaco has been producing VPX boards for a decade now and is fully up the learning curve—but potential customers may not be in that same enviable position. Programs have a long life cycle, and the teams working on them and looking at a refresh are probably steeped in man-centuries of experience with VME. Changing to VPX for the first time can be a big step and one not worth taking if the performance benefits of VPX are not necessary to achieve the program goal.

Cost
In general, VME boards tend to be less complex, by design, than VPX boards and therefore offer a very cost-effective option when considering a hardware update. But: it’s not only the boards that are cost drivers for the integrator—it’s the whole program cost of backplanes, power supplies, chassis and, perhaps most importantly, the software investment required.

Taking all the above into account, and with more emphasis these days on “good enough but quicker, and at lower cost,” many programs are still deciding to go with VME—predominantly for legacy upgrades, but also for new program starts where it makes commercial sense.

In parallel with the bus bandwidth improvements, developments in CPU and I/O technology have also been substantial. Whereas, in the early days, separate boards were required to add Small Computer System Interface (SCSI) and Ethernet capability, current VME boards now support the latest system-on-chip (SoC) devices, offering a single piece of silicon with CPU, memory controller, and I/O all rolled into one.

Three Reasons
VME still has a role to play. Three key reasons that lead to the selection of VME are:

  1. Life extensions
    Programs often have a life that is much longer in reality than was planned for. Airframes and vehicle chassis are valuable assets that tend to find uses long after their original “use-by dates.” Being able to add several extra years to the on-board equipment by upgrading the core VME boards without disrupting the platform yields significant cost benefits for the end user.
  2. Performance enhancements
    As program lives extend, the mission often expands to meet the needs of new technologies or new operational scenarios. Doubling the performance of the CPU by refreshing a five- to seven-year old single board computer enables these mission objectives to be met. An example of this would be adding an embedded training capability to an existing display computer.
  3. SWaP consolidation
    Reducing size, weight, and power usually conjures up ideas of using smaller and smaller modules—but it can also be imagined in the context of compressing the functionality of several existing boards into a single board, usually by use of the multicore technology prevalent on VME boards over the last five or so years. This can leave slots spare in an existing chassis to add extra functionality, perhaps eliminating the space claim of a separate chassis originally performing the added function.

Considering the Full Program Life
Once the requirements for the program are fully understood and finalized and the product is identified, the next objective is to manage the full lifetime of the program. As usual, there are multiple aspects to be considered.

Figure 2: The Abaco 6U VME PPC11A is form, fit, and function compatible with the PPC4A introduced almost 20 years ago.

Technology Insertion Strategies
Being able to slot in new technology with minimum disruption is what technology insertion is all about. Vendors make a choice when designing new boards, weighing ease of design, lowest cost, or absolute maximum performance. However, those vendors  most successful in the VME world focus on minimizing the integration needed by their customers when introducing a new product into their systems. This is particularly true where an upgrade of an existing system is planned.

The core of a technology insertion roadmap is maintaining standard hardware and software interfaces—essentially, keeping pin-outs and APIs the same—or at least maintaining backward compatibility. A challenge for the hardware designer is to bring in new features (such as SATA) when older features (such as IDE) go away and are no longer supportable. The key is to have extremely flexible pin-out schemes with options that offer new features together with options that mimic the pin-out of previous generations. This isn’t easy, as it complicates the design and can be a headache to implement—but it is a necessity to support long term programs where one or more technology refreshes are required along the way.

Design for Longevity
Another key approach is to design-in long life in the first place, therefore minimizing the number of refreshes required. This requires obsolescence and component lifetime issues to be considered up front. Wherever possible, components can be chosen from suppliers “embedded” roadmaps that promise seven, 10, or even 15 years supply. Beyond this, specific action can be taken to remove any uncertainty around longevity of supply of components by implementing the function in IP, programmed into an FPGA. This gives the vendor portability, which does not rely on specific hardware components remaining available on the market. One such example is Abaco’s Vivo VME Bridge, which is on our most recent board designs. As well as implementing core VME technology, the device also manages the high-speed 2eSST functionality and hardware byte swapping that is often required when updating systems with little-endian and big-endian legacies.

Obsolescence Management Strategies
Component obsolescence is a fact of life. Most components are designed for short-lived commercial usage and not for long term programs that represent a tiny fraction of the overall consumer market. Therefore, as well as adopting best design practice, it’s important to have well thought-out obsolescence management strategies. Continuous monitoring of the Bill of Materials (BOM) is essential, but taking the subsequent appropriate action is even more important. The action can be to undertake a Last Time Buy (LTB) or to attempt a redesign that removes the need for the component.

Conclusion
Over its lifetime, the original VME architecture has been substantially enhanced to keep it current and viable. Today, it benefits from the latest silicon technology, the most up to date I/O functionality, and the newest storage capability. Over many years, vendors like Abaco have played a key role in ensuring its relevance with frequent technology refreshes that have enabled customers to regularly leverage higher levels of performance and functionality—at minimum cost and risk.

Given the extensive deployed base of programs that rely on VME architecture, the often multi-decade nature of those deployments, the familiarity with, and trust in, VME that exists widely within prime contractors, systems integrators, and OEMs, and VME’s inherent advantages for certain applications—it is clear that VME will continue to be a significant force for many years to come.


Richard Kirk graduated from the University of Manchester in 1984 with a BSc degree in Physics and followed that in 1998 with an MBA from the Open Business School. In the interim, he’d joined Plessey Optoelectronics, part of one of the UK’s most venerable technology companies. He joined Radstone, located in Towcester, UK—subsequently acquired by GE—in 1999, and now has worldwide responsibility within Abaco’s business as Director, Core Computing.

 

Share and Enjoy:
  • Digg
  • Sphinn
  • del.icio.us
  • Facebook
  • Mixx
  • Google
  • TwitThis

Tags:

Extension Media websites place cookies on your device to give you the best user experience. By using our websites, you agree to placement of these cookies and to our Privacy Policy. Please click here to accept.