Ethernet in the Mix
Flexible, low-cost, and well-understood technology shows up across the board.By Cheryl Coupé, Editor
Ethernet applications run from high-performance data centers and military communications, to smart home devices and entertainment systems. And while the proliferation of mobile devices is pushing wireless into more and more areas, Ethernet isn’t going away. We talked to John Neil Thomson, director of software engineering for GE Intelligent Platforms and Rawin Rojvanit, principal applications engineer, Advanced Microcontroller Architecture Division for Microchip Technology Inc. to get some insight into where Ethernet fits in today’s cloud-based-social-media-Internet-of-things.
Embedded Intel Solutions: What challenges are arising in the connectivity of Ethernet to other interfaces such as PCI Express, VME, Advanced TCA, etc. as those technologies evolve?
Thomson, GE Intelligent Platforms: Ethernet is becoming very ubiquitous—in all form factors and bus types. It has advantages over other connection types in that it is very flexible, scalable (in range and speed), and--of course--low cost. Combined with the fact that its use is easily understood by application programmers, this means that it is now the default for much system design. The idea of the packet-switched backplane (first introduced in cPCI, then adopted by VME and ATCA) is a powerful one, and we’d expect to see any future form factor allowing for Ethernet (maybe 100 Gbps) as one of its backplane interconnection methods.
Embedded Intel® Solutions: What is the role of Ethernet in new smart-grid initiatives in comparison to wireless technologies, such as ZigBee or WiFi?
Rojvanit, Microchip Technology: Many technologies will have a role in the smart-grid initiatives. The design choice will depend upon requirements for bandwidth, security, latency, and physical plant considerations. For higher-bandwidth, higher-reliability, backhaul, secured, service-critical parts of the grid, wired Ethernet is the way to go. Wireless protocols’ strength is for the last 50 m of mobile delivery to end nodes. Wireless-technology choices can be ZigBee or Wi-Fi. The ZigBee protocol is preferred for supporting large numbers of very-low-bandwidth nodes. Wi-Fi is a good choice where access is required to the Internet or by an off-the-shelf consumer platform—that is, either a smartphone, iPad-type device, or platform on a consumer home network. Alternatively, we are seeing more embedded devices that now support both Ethernet and wireless technologies. All of these different technologies ultimately enable devices to communicate with each other. This is a main ingredient in what makes a grid smart—the ability to measure, communicate, and control.
Thomson, GE Intelligent Platforms: We would expect Ethernet to be “in the mix” for most complex, multi-level project designs for many years to come.
Embedded Intel® Solutions: What’s the future of Ethernet compared to WiFi in embedded applications where high bandwidth may not be required?
Rojvanit, Microchip Technology: This question is similar to the previous question. Our response is similar. An additional point is that Ethernet and WiFi are complementary technologies. Non-home setups typically involve both Ethernet and WiFi. Future home networks may also utilize more of both technologies. Stationary applications can conserve WiFi bandwidth by using Ethernet instead. As more devices in the home become network-enabled, balancing the setup between Ethernet and WiFi will become more important.
Thomson, GE Intelligent Platforms: Our customer base and their range of applications are vast, so we’d expect to see Ethernet, WiFi, and future technologies all being used in various ways. Our most successful customers tend to adopt a “horses for courses” approach.
Embedded Intel® Solutions: With new home and video devices pushing the demand for Power over Ethernet (PoE), are there design issues in implementing this technology?
Thomson, GE Intelligent Platforms:Questions are being asked about what power levels can be supported in the future, and we see customers looking for different voltage ranges.
Embedded Intel® Solutions: What kinds of embedded applications will utilize Ethernet in the future?
Thomson, GE Intelligent Platforms: We’d be inclined to ask, “Which won’t?” Ethernet is being used in almost every application field. Even areas that used to be the domain for specific solutions are migrating to Ethernet. Just as one example, many sensor-type devices, which used to sit on RS-485 buses, have now migrated to Ethernet. This is a trend which we are not seeing slow down.
Rojvanit, Microchip Technology: Many devices already utilize Ethernet and the list is growing. Pretty much any embedded application that needs to connect to the Internet could use Ethernet. It is more common to find Ethernet in a stationary device, and a portable device could use WiFi. In terms of security, some applications may prefer Ethernet, as the network cannot be easily sniffed unless someone has direct physical access. In general, protocol security should be the main defense for data security and privacy. Some applications may also prefer Ethernet’s ease of setup, as it doesn’t require a password. High-bandwidth embedded applications are likely to continue depending on Ethernet.
Embedded Intel® Solutions: As IP version 6 is deployed, what compatibility issues with version 4 should device developers be aware of?
Rojvanit, Microchip Technology: We expect both IP version 6 (IPv6) and IP version 4 (IPv4) to coexist for some time. The multitude of existing IPv4 equipment will not become obsolete overnight. New embedded products that support IPv6 will most likely support dual IPv4 and IPv6 stacks, keeping the levels of interoperability high. It will be difficult for developers to predict when end customer and ISP networks will be upgraded to support IPv6. Networking equipment and infrastructure will be upgraded to support IPv6 over time, but new equipment will likely still support IPv4. There could be some reachability issues if a device is IPv4 and wants to reach a server with an IPv6-only address. Having dual IP stacks would maximize interoperability over the long term.
As the migration to IPv6 occurs, there will be dual backbones and stepped changes to the routing infrastructure. A forklift change to the worldwide network is not about to happen. The primary difference between the two protocols is also likely to be the point of contention for compatibility on the migration. This difference lies in the addressing and routing technique that is used. Thus, as developers consider the migration to IPv6, consideration must be given to the limitations of both protocols to expectations of the other. This will apply to addressing (e.g., greater resolution in IPv6) as well as service identification (e.g., greater direct support in IPv4).
Thomson, GE Intelligent Platforms: We have been supporting IPv6 in our product ranges for many years now, and have seen customers roll out IPv6 networks. Various types of tunneling and co-existence strategies have been used. These can be complex, and we would suggest that careful planning is necessary. Seamless migration is possible—but probably not easy.
Cheryl Berglund Coupé is editor of EECatalog.com. Her articles have appeared in EE Times, Electronic Business, Microsoft Embedded Review, and Windows Developer’s Journal. She has developed presentations for the Embedded Systems Conference and ICSPAT. Coupé has held a variety of production, technical marketing, and writing positions within technology companies and agencies in the Northwest.