As one of the hottest areas of engineering activity, the networking and communications segment of embedded design has many new standards emerging. Hardware and software designers must be aware of the details of technology evolution while being active in their development.
This week's In Focus looks at how designers from one company Motorola deal with product development – integrating basic technology, business priorities and the critical issues of standards. These engineers are charged both with developing successful product strategies and with playing a key role in the evolution of industry standards. They are in a unique position to gauge how effectively they and their colleagues' engineering efforts converge on emerging technology areas.
Some standards have become traditional standbys such as VME and cPCI while others are new; Infiniband, Starfabric, and RapidIO to name a few. Others are software-oriented and emerge out of industry groups such as the Network Processing Forum (NPF), the Multiservice Switching Forum (MSF), and the Service Availability Forum (SAF).
But, while the usefulness of standards is not new, what is unprecedented is the dependence in broad segments of the industry on the creation and development of commonly agreed to standards based on broad-based consensus and cooperation. This is a different concept than ongoing battles between companies to establish their proprietary spec, either in the market or within an industry group, as a defacto standard.
In the net-centric computing and communications era, emerging standards are more and more derived from a process of detailed technical study, consensus and negotiation.
What has happened in the adoption of television, radio and the Internet offers a view of the impact that standards will have on the deployment of the embedded networking infrastructure. These industries have all gone through adoption cycles that took years before they were accepted by the masses. Standards are what made them practical and cost effective for mass adoption.
In the beginning, everyone had their own custom way of implementing the technology, but as products evolved, key standards developed that made the technologies successful. “I think that what has become clear to almost everyone in this business is that we can't let this process of standardization happen organically,” said Bruce Rosenkrantz, a contributor to this week's In Focus and a distinguished member of the technical staff at Motorola's Computer Group (Temp, Ariz.) who serves as chairman of the PICMG 2.20 standards subcommittee, “because that is just too slow and unpredictable. We have got to make it happen through on-going participation in real industry-wide standardization efforts.”
What are the standards evolving in this particular segment that developers will have to learn? And out of the creation of these standards what rules and procedures can be derived to help in the development of standards?
According to Jerry Gipper, director of marketing at the Motorola Computer Group, the answers to the second question are no less important than the first, perhaps more so, because the new network-oriented computing and communications era is the child of the democratic and consensus driven standards methodologies that evolved out of the Internet.
“As virtually every computing device is now becoming connected, the importance of the embedded networking segment cannot be overstated,” he said. “In terms of consumer usage of computers, the shift is from a desktop paradigm to one in which there are several more personal and connected options available centering around a variety of smaller netcentric information devices: Web-enabled PDAs, Internet-capable wireless phones, DSL-capable wired phones, and cable modem-enhanced set top boxes,” he said.
He believes that the embedded network infrastructure that is needed to support this kind of net-centered computing has to be built as quickly as possible. Gone, he said, are the days when an industry can wait for a defacto standard to emerge from competition among several large contenders. And, while the process would more likely result in true industry wide standards, gone also are the days when standards could be developed slowly in a long process through various national and international standards and professional bodies.
According to Gipper, groups of companies are gathering together long before hardware and software solutions are needed to reach some sort of consensus, first on an informal ad-hoc basis and then in more formal market and application specific trade organizations. “In a connected environment, it is clear that no one company can go it alone and hope to impose a particular proprietary solution,” he said. “In such an environment, if a company chooses to go its own way it's products will not work with anyone else's.” Such companies, if they do not participate in true standards efforts, will quickly find themselves on the outside looking in.
In the past year or so, in particular, he said, standards activity has accelerated across the board, not only in diversity but in the speed with which problems and conflicts are being resolved. “Older standards efforts in embedded networking such as VME and compact PCI are accelerating their pace and new standards efforts are forming and coming to consensus much more quickly,” said Gipper. “And efforts that previously might have taken several years are moving from definition to proposal for formal approval in nine to twelve months in some cases.”
According Rosenkrantz, chairman of the PICMG 2.20 subcommittee, the alphabet soup of new serial interconnect standards that are being developed to link devices in the embedded networking infrastructure may be baffling to the uninitiated. “But standards groups have taken great care to make sure that each one performs a needed job particular to an individual application,” he said. And while it would appear that there is a great deal of overlap and the potential for conflicts between these standards, it is more apparent than real.
“The conflict is largely on the marketing level in which some companies backing a particular standard are looking for the broadest possible number of markets to lower costs, even though they are outside the parameters of the original specification,” he pointed out. “At the technical level, however, conversations and interactions between the various standards groups are going on all the time. While a lot of this interchange goes on informally, more and more there are efforts to formalize such exchanges, to ensure that technical mismatches do not occur.”
One area where this has happened is between the standards groups focused on existing parallel shared bus topologies such as PCI, compact PCI and VME and companies that are migrating from current architectures to the newer higher speed serial interconnects. As described by Jeffrey Harris, director of research and system architecture at Motorola, one effort to provide an environment by which such migration can occur is the VMEbus Switched Serial specification under definition by the VMEBus Standards Organization.
To provide the nexus where embedded hardware and software standards and specifications meet, is a collection of industry groups whose roles have taken on increasing importance in embedded networking applications. These include the Network Processing Forum (NPF), the Multiservice Switching Forum (MSF), and the Service Availability Forum (SAF).
In addition, developers who want to keep an eye on where embedded networking is going, will have the services of a half a dozen industry groups, including the ATM Forum, the Voice and Multimedia over ATM (VMOA) working group of the ATM Forum, the European Telecommunications Standards Institute (ETSI) and its TIPHON (Telephony IP Harmonization Over Networks) working group, and the Third Generation Partnership Project. Members of some of these groups describe their missions in this section and the progress they are making toward standardization.
“If there are concerns that these standards and specifications may conflict with one another in any way,” noted Gipper, “one way to ensure that does not happen is for engineers in all areas of embedded networking to get actively involved as members of such standards organizations.” The more working engineers and developers involved, he said, the more assurance that there will be cross-fertilization enough to deal with such conflicts before they get serious.
And, if there is any area of embedded networking that standards groups have not addressed fully, said Henry Wong, principal staff engineer at Motorola and an active member of both the PICMG and VME Standards Organization, it is optical interconnect. “With the travails in the networking and communications segments, the speed with which optical will replace copper has slowed down to a crawl in the general wide area environment,” he said. “But with time, this segment will recover and companies will start thinking about moving optical down into systems as the connection of choice between boxes, then between boards, on the boards and ultimately, on-chip.”
While there are always unknowns that will emerge as this evolution occurs, Wong feels the standards efforts of all of the major embedded networking groups have factored optical into the equation. “In current switched fabric standards, such as Infiniband, RapidIO and various PCMIG and VME efforts, the logical protocols and the interconnection schemes will remain unchanged,” he said, “with the actual physical connectors being the only thing that would have to be modified.”