RTC Magazine

The PCI Industrial Computer Manufacturers Group (PICMG) introduced CompactPCI or PICMG 2.x in 1995. At the time, it represented a sea change for the embedded computing industry. The form-factor offered a valuable new choice to the industry, which had previously been dominated by the VME bus standard. CompactPCI quickly became widely adopted, particularly within the telecommunications market. The technology anticipated and accommodated leading embedded computing technology innovations, including high-end processors, high-speed bus backplanes and state-of-the-art power delivery and cooling.

Recently, an increased number of COTS military and government applications have been migrating to CompactPCI (cPCI) as a viable pin-and-socket follow-on to their existing VMEbus systems. The migration is being driven mainly by performance, availability and the ease of implementing contemporary communications technologies. Some analysts have forecast that further penetration into applications including the defense and homeland security market will cause cPCI to experience further growth.

Over the last 11 years, cPCI has undergone a continuing advance of “point specs,” which have incrementally improved the architecture, adding features such as hot-swap, redundancy, telephony bus and support for telephony mezzanine modules. Most dramatically, in 2001, PICMG 2.16 was developed to extend the life of existing cPCI systems by combining the inherent robustness and reliability of cPCI with packet-switched Ethernet fabrics. So popular was this new development, that within 18 months of its ratification, more than 40 companies were producing and shipping PICMG 2.16-compatible products.

PICMG 2.16 eliminated bottlenecks in the traditional bus approach taken by VME and original cPCI technologies. Based on a network of independently switched nodes, it not only removed the limitations on the number of system elements, but also on the overall throughput capacity of the system. The PICMG 2.16 specification describes a redundant, switched 10/100/1000 Ethernet network within a cPCI chassis, providing IP connectivity using a “star” topography routed across the backplane for an aggregate bandwidth of up to 40 Gbits/s.

This specification allowed for flexibility in selecting the connection speed and higher-level protocols based on specific application need. This frees equipment manufacturers to mix and match switches, single board computers, chassis and sub-components. The modular make-up of PICMG 2.16 system design simplifies deployment, increases serviceability and greatly improves reliability. In addition, having the switch in the chassis minimizes external wiring (Figure 1).

VME - Over the hill?

There is no denying that VME has an impressive lifespan. It has recently celebrated its 25th anniversary as an open system architecture, has continuously evolved its core technology and has introduced several advanced interconnect and fabric capabilities. According to some analysts, VME bus technologies still account for over 25 percent of all embedded market applications.

However, with cPCI entering the market in the 1990s, and now the recent rise and growing acceptance of AdvancedTCA and MicroTCA in the telecom space, many are left wondering—has VME been taken as far as it can go through incremental improvement? Should it be phased out in favor of other form-factors that are designed from the beginning to accommodate the technology infrastructure as it exists today or is anticipated to require in the future? Will the components remain available over time? And will the ecosystem of suppliers continue to shrink?

PICMG 2.16 is a mature, well-proven, deployed technology used in production today by hundreds of customers spanning many industries, from telecommunications to defense and homeland security. Experience has shown that the one thing embedded market customers look for is the size of the ecosystem. No one wants to use “standardized” products only to find that the “version” they want to use is not supported by many vendors. By far, the biggest advantage that PICMG 2.16 has going for it is that its vendor base is quite large, and that interoperability is a given, because there are really no choices to be made within 2.16. All other standards, including the newest from VITA and PICMG, have so many choices that it is extremely difficult for a customer to find more than a few vendors or products that are truly compatible.

As this magazine’s editor recently mentioned, the networking industry is rife with examples of technologies that were designed for one purpose and wound up used and modified for other, equally lucrative applications far from the original intention.

cPCI began with a telecom base both in central office and remote sites, but has grown to encompass military, industrial, and other governmental applications. In particular, with the rise in military budgets since the early 2000s in both European nations and in the United States, cPCI has come to represent a major threat to established bus architectures including VME, especially within new military and defense projects looking to implement a more contemporary IP-based architecture.

Governmental and industrial equipment integrators have the same fundamental needs as telecom equipment manufacturers and have been moving to cPCI systems for fixed applications including factories and remote sites, as well as for mobile communications needs including airborne, shipboard and military vehicles (Figure 2). In these scenarios, VME-based systems struggle under the weight of throughput limitations due to their bus-based approach.

A number of new technical approaches have been tried in VME to staunch its loss of new design activity and market share, but as often happens, introducing new technologies like newer, high-speed interconnects for VME often leads to market fragmentation. Multiple choices have to be made between new quasi-proprietary buses and a number of networking topologies. In turn, such a plethora of choices leads to a confused vendor and customer base and limits the number of suppliers that an integrator can call on for solutions using the particular topology he or she has chosen.

In contrast, systems based on Ethernet, Gigabit Ethernet and now 10 Gbit Ethernet (which is supported in 2.16 environments, but not in the backplane) have come to dominate cPCI, leading to a focused, mature ecosystem and broad choice of competitively priced vendors who offer an array of compatible blades. For example, there are now more than thirty suppliers of single board computers for the cPCI 2.16 market. Integrators are able to choose from a wide selection of competing boards that cover everything from low-cost Intel Pentium solutions to high-performance PowerPC and dual core x-86 solutions, along with a very impressive array of real-time and standard operating systems and applications for those blades.

Rugged Designs and Remote Management

It is easy to see why military and defense and homeland security are growing markets for cPCI as a migration path from existing VME COTS systems. Equipment vendors need to ship equipment that is tough and reliable and built to withstand extremes in temperature, shock and vibration.

Due to its original market focus, just about all cPCI products are built to satisfy the rigorous requirements of the NEBs Level 3 telecom standards. This means that just about every board can, in fact, handle the Mil-Spec requirements dictated by government programs. Even if they don’t meet the requirements at the outset, most can be readily modified to do so. Conformally coated, ruggedized and conduction cooled, cPCI blades and systems have been deployed for everyday use worldwide by prime contactors and governmental users.

Since they deal with a large number of remote locations, military and defense system managers must be able to remotely manage that equipment for software upgrades, alarm conditions and system control. Thanks to network-based blades and system management modules common in cPCI systems, they can easily do that.

Network-based systems often use commercial packages based on Simple Network Management Protocol (SNMP), File Transfer Protocol (FTP) or its secure cousin SFTP, to remotely control and update individual system elements via an open or secure network.

Similarly, the chassis itself can be completely controlled via an out-out-band connection to the PICMG 2.9 Intelligent Platform Management Interface (IPMI) controller or set of redundant controllers, which act as the single management entity responsible for overseeing all of the boards, power supplies and fans within a system. The IPMI system manager can automatically react to stimuli such as an over temperature condition, a board watchdog event or power supply failure, while at the same time sending an alarm to a central site. Many 2.16-based systems are specifically designed to shunt system workload to an alternative resource, either within that chassis or to a hot standby system (Figure 3).

Keeping up with the Joneses

Backwards compatibility is another major feature required by military system designers. The military has been using embedded systems for more than forty years. Unlike the PC market, where everything changes every nine months, maintaining and refreshing military systems, which are often based on custom designs with frozen technology, is a major challenge. The OS and drivers used by all boards in a bus-based system must be locked together, creating another impediment to keeping designs fresh and being able to integrate the latest technologies and techniques to adjust to user needs.

Using 2.16 COTS systems changes that paradigm. A switched packet backplane approach grants a much finer granularity to system changes, since the now independent system nodes use the lingua franca of Ethernet to communicate at the network level, severing the master/slave co-dependency of bus-based systems. In this new scenario, the user simply changes one single board computer or communication node and puts in a higher performance unit to upgrade the whole system, greatly lessening the impact of obsolescence and upgrades.

cPCI 2.16 represents the best, most balanced solution, because regardless of individual connection speeds, the fundamental technologies, topologies and protocols all work together. It is a standard that is evolving over time, but is inherently backward compatible. Elements of switched Ethernet backplane systems may all talk at different speeds, but all speak the same language.

To some degree, PICMG 2.16 can attribute its success to its ability to leverage the ubiquity of Ethernet. Since the future of embedded application development is all about IP and to a large extent IP is so directly associated with transport over Ethernet—95 percent of the world’s data rides over Ethernet at one point or another—it only makes sense to build your infrastructure natively with Ethernet. Therefore, 2.16 has set the stage for this “revolution,” with other specifications including AdvancedTCA and MicroTCA following in its footsteps.

Industry analysts support the idea that companies will still require cPCI even as the new AdvancedTCA and MicroTCA specifications become more widely accepted. For example, some analysts forecast that further penetration into applications including military and defense will cause the cPCI market to experience strong growth over the coming years. Although VME still has its place in legacy systems and AdvancedTCA solutions are an excellent choice for core telecom or other high-performance applications, growth in the cPCI market will continue for more and more applications that do not require the outright performance—and associated price and footprint—of AdvancedTCA.

Performance Technologies
Rochester, NY.
(585) 256-0200.
[www.pt.com].