Whole Program Life COTS: warding off the curse of component obsolescence

Component obsolescence is the curse of COTS (commercial off-the-shelf) technology.

By Peter J. Cavill

Component obsolescence is the curse of COTS (commercial off-the-shelf) technology.

There has never been a more demanding time for developers of aerospace and military systems destined to operate in some of the harshest environments — on or off the planet. With semiconductor manufacturers releasing improved processors typically every three months and new generations every 18 months, system designers have access to more processing power at a lower cost than ever before.

Increasingly, however, the specter of component obsolescence is an unwelcome presence at the feast. Today, where the needs of the desktop rather than the aircraft cockpit drive the pace of innovation, there is a complete mismatch between military system lifecycles and the COTS components from which they are built. While commercial markets can plan — and benefit from — new systems every three years or so, military programs tend to live for 15 years and more.

In response, experts at Radstone Technology have begun a new program called Whole Program Life COTS to mitigate the effects of component obsolescence in Radstone's single-board computers and computer subsystems. This program aims to reduce overall cost of ownership and provide industry-leading safeguards against obsolescence. It does so by addressing the issue at all stages of the program life cycle, from designed-in compatibility across generations of products through to a team of design engineers and support staff dedicated solely to supporting programs throughout their 15 to 20-year lives.

In developing an effective response to the reality of component obsolescence, it is important to understand the nature of the problem. There are, essentially, three types of obsolescence.

The first is obsolescence caused by technological evolution. This is where a new generation of technology is developed which effectively makes its predecessor obsolete. Typically the new generation will improve performance and functionality, often at lower cost than its predecessors will. Importantly, however, the new generation will typically perform the same function as the previous one.

The second is technological revolution where a new technology emerges that supersedes its predecessor. An example of this is the Fiber Distributed Data Interface (FDDI) that is becoming obsolete as the market moves towards adopting Fibre Channel as the communications technology of choice.

The third is obsolescence that market forces cause. This is where demand for a component or technology falls to where manufacturers consider it uneconomic to continue production. This is an increasing problem, as the military and aerospace market no longer commands the purchasing power necessary to persuade manufacturers to continue production, for example, of a particular processor.

As the sensitivity to obsolescence varies enormously from system to system, so does the range of responses required. Radstone's experience shows there is no one 'silver bullet' or cure-all (despite what some might profess). Instead we have developed our Whole Program Life COTS approach as a coherent design philosophy, combined with industry best practice and expert dedicated support resource, which seeks to protect developers from the pain of obsolescence. The aim is to provide the most appropriate response at any particular point of the program lifecycle.

Radstone's preferred route towards dealing with obsolescence is technology insertion where we take responsibility for ensuring that successive generations of products are fully backward compatible with the previous generation.

The starting point in achieving such continuity is to have a good understanding of the silicon roadmap being followed by key semiconductor manufacturers. Radstone benefits from a close relationship with its key technology suppliers, including access to product road maps and early access to the latest products.

Standard board architecture

One example of this is Radstone's range of PowerPC boards. When we developed our first PowerPC board in 1995 we took chose a standard architecture for the board in terms of hardware and software. This architecture was largely independent of the 30 or so key silicon components used and which we considered to have a long-term future.

Standardizing on the PowerPC Reference Platform (PReP) architecture has enabled us to change processor and memory types and speeds without changing any of the physical or electrical characteristics of the board. By consistently maintaining the same form, fit and function Radstone has been able to deliver truly transparent 'plug and play' technology insertion through five generations of processor technology.

It is possible to remove a board running at 100MHz that was manufactured in 1995 and replace it in the same slot with our latest 500MHz board. It will run the same application code but with a tenfold improvement in performance. Importantly, this has been achieved without the need for our customers to carry out even minor system re-engineering in order to accommodate technology insertion.

The last 12 months or so has seen the emergence of the Common Hardware Reference Platform (CHRP) architecture as the successor to PReP. Radstone has already launched a new range of CHRP-based boards for embedded applications using Motorola's 8240 series of processors. Importantly, CHRP is also on the latest generation of Radstone's PowerPC boards — the PPC4B. But even with a change of architecture, Radstone is still able to offer backward compatibility through its software strategy.

Radstone supplies low-level software that enables the feature set of Radstone hardware within the context of standard COTS software. Radstone's software is continuously evolved to ensure compatibility with the latest releases of the market-leading COTS real-time operating systems.

Radstone software conforms to the COTS philosophy; it is modular, operates on a wide range of Radstone hardware, incorporates mainstream standards, and is certified through stringent testing.

The overriding aim of Radstone's software is to insulate the operating system and user application from low-level hardware issues. We do this through a series of software layers, or filters, that mask the application software from technology changes whilst enabling users to benefit from increased performance.

SilverChip technology, which also forms the basis of Radstone's VxWorks bootroms, filters out differences in memory, host bridge, and Level 2 cache technologies, as well as addresses many peripheral device and interrupt routing issues. It allows Radstone to take advantage of component functionality or performance improvements without disturbing either the user's or Radstone's BSP.

For example, engineers implement some bus bridge devices as completely different devices on Radstone boards within the family — partly due to chip upgrades and partly due to environmental constraint issues. SilverChip initializes the 'different portions' to make the devices look the same when the operating system sees them.

SilverChip also performs PCI configuration including all nodes, multi-function devices, as well as memory and I/O space allocation. The PCI configuration function, in essence, 'shields' operating system drivers from system configuration issues.

Military and aerospace system developers face the dilemma that whilst they wish to take full advantage of the performance and feature improvements of COTS technology, there is the equally important need to maintain system stability over long periods.

Increasingly, Radstone is working with system designers to take advantage of such progress through technology refresh at key stages of a program. Typically opportunities occur at each major tranche of a long-term production program, perhaps at five- or 10-year intervals.

Radstone's commitment to maintain form, fit, and function through careful design and software layers can effectively extend the life of a system for 10 to 15 years without any major re-engineering. Any changes to the underlying silicon will have been hidden from the customer's application.

Long-term management

Beyond 15 years, strategies for dealing with component obsolescence revolve far more around management than they do around technology. A fundamentally important element of Radstone's Whole Program Life COTS is the various ways in which we address the issues of long-term support.

Radstone leaders have come to the conclusion that in-life support can only get the priority it deserves through a dedicated focus. It has, therefore, created a Product Lifecycle Management (PLM) organization that enjoys dedicated engineering resources of the level normally given to new-product design teams.

Dedicated component engineers maintain close relationships with vendors to ensure that Radstone remains informed about the availability of components. They ensure that the team always has the in-depth knowledge and experience of the marketplace necessary to establish sources for more difficult-to-acquire and end-of-life components. And, as a member of the Component Obsolescence Group, Radstone is able to keep abreast of industry-wide developments and innovations in this field.

In addition to being offered the opportunity to make 'last-time buys', customers can also make a 'one off' purchase of all the components that will be required throughout the entire life of a program.

As part of a long-term support agreement, Radstone will hold stocks of components as a customer's property, in a secure and environmentally controlled 'bonded store' facility.

But this cannot be the whole solution; what of the ongoing need to identify, trap, and act upon component end-of-life that are being issued all the time? That is where a Radstone 'health check' comes in. The health check comprises a quarterly audit by a component engineer on the availability of all parts that make up a particular product.

This helps manage the risk of ownership by minimizing the danger of components becoming unavailable and maximizing the opportunity to take appropriate action. Health checks can be arranged, on an annual basis, at any time in a product's life cycle.

The ability to understand a customer's system, and the interaction of one component or sub-system on another, is critical in today's world of shortening silicon lifecycles. Radstone dedicates hardware and software engineers to technical support. There is no time wasted waiting for access to the same development and test equipment that the product's original design team used: the PLM organization has its own suite to ensure that bottlenecks and conflicting cross-divisional priorities do not occur.

This strategy is being used by Northrop Grumman to safeguard its Firefinder tactical battlefield weapon program. Firefinder is able to 'see' incoming hostile fire, identify the type of projectile being used, and then calculate the point of discharge so that friendly artillery batteries can destroy the enemy weapons site.

Firefinder was first introduced in the 1970s. In the early 1990s Northrop Grumman won the contract to update the system and Radstone subsequently supplied its VSP-1 vector signal processor platform; CPU-44 68040-based CPUs; and PIO-2 parallel I/O boards, along with appropriate mechanical and thermal design re-engineering.

Field trials and qualification followed, as did early tranches of production systems. However, nearly a decade later, a number of key components have become obsolete. Faced with the requirement for more modules, Northrop Grumman in August 2000 commissioned Radstone to combat obsolescence by re-designing the printed wiring boards to accommodate the different packaging and associated circuitry used by today's generation of silicon, and re-targeting the design of custom ASICs to today's technology.

Peter J. Cavill is managing director of Radstone Technology, a designer of ruggedized single-board computers and computer subsystems based in Towcester, England, and is responsible for the Radstone Group's Embedded Computing business.

More in Computers