By Cheryl Coupe
As challenges such as security and multicore processing are addressed, Android and Embedded Linux find their way into new markets, moving from traditional smartphone applications to medical, automotive, mil/aero and M2M of every flavor.
EECatalog: Android’s biggest appeal in embedded is its combination of Linux underpinnings, huge commercial momentum in smartphones, and off-the-shelf GUI support. What new embedded markets is Android finding its way into?
Karthik Ranjan, ARM: Traditional embedded markets have historically been a mix of custom Linux solutions combined with Windows Embedded Compact (CE) operating systems. Android is finding its way into a range of both consumer and enterprise devices not only because of the aforementioned value and momentum but also because of the appealing royalty-free business model. This has enabled it to penetrate new markets such as set-top boxes and digital televisions such as those found in China— for example Geniatech and TCL. Android is also finding its way into the new emerging smartwatch category such as those from Italian-designed I’m Watch. However Android is not just limited to consumer electronics; it’s also finding its way into traditional enterprise strongholds such as the lucrative enterprise handheld terminal market. Once a market dominated by Windows Embedded, traditional vendors such as Honeywell and Motorola Solutions have all adopted Android-based solutions.
Willard Tu, ARM: As mentioned in the question, Android is getting more adoption in areas that leverage its off-the-shelf GUI support. Lots of embedded devices are considering or adding a graphical display. I can think of medical instrumentation, toys, scanning devices and digital signage—many of these are not high-volume apps that can warrant creation of a custom Linux solution, hence they try to purse something that already exists such as Android, which in many ways fits the bill and is why it is getting a lot of consideration or adoption.
Chris Buerger, Wind River: Android continues to provide a flexible option across numerous embedded opportunities. Some of the most active areas continue to be aerospace and defense, automotive and medical device applications.
Android is steadily gaining traction in areas that require rich graphics and/or increasingly need multiple OSes on one device…imagine a device that must include an RTOS but also Android for graphics-heavy UI. For example, we’re also seeing the community that previously used Windows Embedded now looking for a change and investigating Android.
Keep in mind that open source is already well-adopted in the mission-critical and carrier-grade arenas of networking such as core and edge devices. Also, it is growing significantly in strict security-related applications and it’s evident by the significant adoption of Linux in military, aerospace and government.
There are mixed industry views of using open source in safety- or mission-critical applications. While Linux and Android are not designed for applications requiring higher levels of safety and are not formally certified, there is an ongoing effort to grow acceptance of open source in safety-related systems which could provide a sound basis for building safety systems in the future.
Warren Kurisu, Mentor Embedded Software Division: We are seeing a fast-growing demand for Android apps in the car as one example. Consumers want to repeat their smartphone experience when driving and this is causing an evolution of new embedded software architectures for in-vehicle use. One of the concerns car makers have is security, and the potential risk caused by introducing “untrusted” software apps into the vehicle. Mentor Graphics has embedded solutions that allow Android operating systems to run separately from core vehicle functions based on Linux or any other operating system.
Incidentally, some car makers have opted for “native” Android infotainment solutions, making use of the full Android software stack. Implementers will normally customize the UI layer, and restrict app support to trusted and approved in-vehicle applications. Android out-of-the-box is not well-suited for this type of in-vehicle infotainment (IVI) use.
On the flip side, Linux advantages include its performance, its community development, its massive amount of available code packages, the GPL and its ability to give vendors control over their destiny. For the embedded market, how do you compare Linux to Android, which some argue is too tightly controlled by Google?
Ranjan, ARM: For consumer electronics, proximity to Google is a clear concern and has prevented the growth of Android into CE devices in U.S. markets for devices like set-top boxes for example, where operators prefer embedded Linux over Android. This primarily resulted in the concerns about a potential conflict of interest with Google over products like Google TV in the most profitable U.S. pay-TV market. However, across the pond in Europe where pay-TV is somewhat less profitable, there have been fewer concerns about using Android for set-top boxes from operators like Swisscom, for example. Another major example of this is the adoption of Android by Amazon for its Kindle Fire product line, where Amazon has been more than happy to replace the default Google Play Store with Amazon’s own App Store.
Tu, ARM: As I mentioned earlier, some embedded devices just do not have the scalability to involve investment in creating a custom Linux distribution. These markets want to leverage what is existing and not re-invent the wheel. Off-the-shelf IP is highly desirable. The only real issue is not control but fit. If Android can meet the application needs, then it makes sense; otherwise, the developer is likely to look to open source to find something they can modify to meet their needs or commercially to acquire.
Buerger, Wind River: The use case and requirements for a project will determine when Linux or Android may be the most appropriate option. For example, Wind River Linux is a very well-established offering and has all the proven proof points and requirements for a commercial Linux distribution. Keep in mind that Android can be thought of as a type of Linux. One reason customers are taking a strong look at Android is that it is a solution stack with a tremendously rich ecosystem of applications that you don’t get with other Linux distributions worldwide.
As a commercialization partner, we don’t treat Linux or Android differently—we’re mainly interested in helping customers find a solution that best fits their needs. Wind River has deep expertise in open source, regardless of whether it’s embedded Linux, Android or others.
Kurisu, Mentor: Linux has over 100 variants, and is truly an open source solution with strong community support. A specific open source community project for vehicle infotainment called the GENIVI Alliance was established in 2009 to allow designers to leverage Linux core capabilities for in-vehicle infotainment (IVI) systems, and the organization now has around 170 members. The first GENIVI Linux infotainment platform goes into production later this year. A true open source platform such as Linux has lower risk for a car maker as compared to one which is managed by a commercial organization (for example Android and Google, Inc). No one is saying that Google will change the licensing model of Android, but the frequency of releases (Key Lime Pie is due next year) is entirely under their control, and car makers will be obliged to absorb any changes in license terms or technical content if they want to keep their designs in step with the Android releases.
From security issues to data-exchange protocols; low-power requirements and new types of CPUs/MCUs—what affect is the rise of M2M computing having on Linux?
Tu, ARM: M2M computing has many new requirements. Security is certainly one of them. How do you secure data in the mesh network, or from the mesh network to the gateway to the cloud? Data is becoming much more valuable and in the past, M2M solutions were based on proprietary networks which were closed loops. Now M2M wants to leverage the Internet. There will need to be development of M2M standards that will help provide security. Currently, I think businesses are still trying to figure out how to effectively create IoT and M2M business models such as software as a service. Security will come to the forefront, once people figure out how to make money.
Kurisu, Mentor: Linux is continuing to establish itself in new-found roles in leading-edge solutions, as implementers look to upgrade from the traditionally more predictable RTOS-based solutions. As the scope of M2M grows, such as vehicle-to-vehicle automated communication, the demands on the computing platform also increase. Linux is a truly scalable, multi-tasking operating system that can take advantage of multicore high-performance SoC platforms and support an extensive array of peripheral devices that exist in a multi-machine environment. Implementers are being forced to consider security and safety issues more carefully; traditionally a potential weak spot in Linux systems.
Samsung just announced that some versions of Galaxy devices—likely tablets at first—can be equipped with Green Hills’ INTEGRITY Multivisor for Trusted Mobile Devices. In effect, this is an NSA-quality partitioned operating environment now baked into a commercial off-the-shelf device. Do you see this kind of architecture as becoming important on Android- or Linux-based embedded platforms?
Ranjan, ARM: ARM is very pleased that Samsung has chosen to adopt the use of ARM TrustZone to be part of the Samsung Knox solution. Combined with other security capabilities such as Hypervisor, TrustZone provides a strong secure execution environment for securing a variety of services on a mobile device including DRM, BYOD and payment solutions. Trustzone can equally be leveraged in the same way to provide solutions for DRM in STBs, secure payment for mobile point of sale (PoS) solutions, as well as enhancing security in traditional embedded enterprise applications such as handheld terminals, industrial automation, etc.
Buerger, Wind River: The concept of embedded virtualization and partitioning is gaining interest. Especially with the rising adoption of multicore chipsets, the use of Linux and other operating systems in a dual-architecture are ripe. In this scenario, it is important to match the characteristics of the OS to the right task. Furthermore, it also allows the reuse of legacy code while using newer operating systems for new development.
As open source continues to grow, especially in areas historically reluctant to use it (such as A&D, auto, industrial and other areas sensitive to safety-critical requirements), we’ll continue to see further experimentation of multiple OSes and partitioning in order to innovate faster while staying on budget and schedule. With multicore and hypervisor technologies, Linux can be used alongside an RTOS in an overall mission-critical/safety-related system. You can “contain” Linux in a non-safe partition and use an RTOS for safety applications, making the overall system safe.
Kurisu, Mentor: Software developers are looking for novel architectural solutions that allow trusted and untrusted domains to co-exist. Separation can be managed at a hardware level, or further up the stack using core separation, firewalls, Linux containers and hypervisors. Mentor Graphics believes it is important to have a range of options available, allowing the implementer to select the right option for each particular job. By consolidating different operating systems onto a single multicore SoC, manufacturing and total system component costs can be reduced.
What affect does multicore processing—including heterogeneous multiple cores—have on embedded Linux development?
Tu, ARM: Multicore processing brings a level of complexity to embedded design. The majority of embedded designs are MCU oriented, and the majority are probably single-core designs. There are multicore devices in MCU where you see a combination of Cortex CPUs. For example Texas Instruments has a family of TMS570 devices that might have dual Cortex-R4 working lock-step for safety applications. Additionally the TMS570 family can offer Cortex-R4 and a Cortex-M3 for other designs that want to leverage the Cortex-R4 for its computational power for algorithm processing, and the Cortex-M3 for its command and control for more interrupt handling. Similarly, NXP LPC4000 devices which pair a Cortex-M4 and a Cortex-M0, where the Cortex-M4 is better suited for some digital signal processing for audio or motor control applications, and having the Cortex-M0 offload the Cortex-M4 and handle other I/O-intensive operations. There are many more multicore processing designs that may use big.LITTLE implementation. In these more application-centric markets, Linux is certainly a favorite. ARM big.LITTLE, is about energy saving. It combines a high-performance CPU with a more energy-efficient CPU so that the application can decide seamlessly which processor to access to optimize user performance.
Buerger, Wind River: The trend and promise of multicore is finally catching real momentum and is a key consideration in all types of projects across the embedded industry. With the power and cost savings that come with consolidation, multicore technology provides companies with capabilities to meet their continued need for faster time-to-market and reduced material/development costs. Multicore will continue to be a factor in most new projects moving forward.
To recap, open source continues to gain traction in areas with safety- and mission-critical applications, such as in A&D, automotive, industrial and medical markets, especially as multicore technologies are maximally leveraged. With an increased adoption of multicore and hypervisor technologies for multiple OS scenarios, Linux can be used alongside an RTOS in mission-critical and safety-related systems through the use of partitioning technologies. Additionally, as the use of multicore and embedded virtualization grows, there is an increasing need for powerful multicore-conscious tools.
Finally, given the added complexity that comes with multicore (multiple OSes, connectivity, richer applications, graphics, etc.), companies will continue to require assistance from expert commercial vendors, such as Wind River. Not only can companies turn to Wind River for its unique blend of deep embedded, open source and vertical-industry expertise, they can rely on our global support and long-term maintenance. While multicore presents increased software complexity, resulting in an entirely new set of challenges, the potential that it offers is undeniable.
Kurisu, Mentor: We are seeing multicore processing in which Linux is one component of a heterogeneous multicore architecture. An example of this is Freescale’s Vybrid Controller Solutions. In this scenario, a robust OS like Linux runs on the more powerful ARM core and the smaller footprint of an RTOS, such as Nucleus, runs on a core where memory, power and overall resources are more constrained. Obviously, having a common communication infrastructure that can be leveraged between OSes is critical. Mentor Graphics has specifically invested in Sourcery CodeBench and Sourcery Analyzer to help implementers take full advantage of multicore processing. Mentor offers a compatible toolchain and debug environment necessary to optimize the distribution of load between different cores, and also debug the performance of a multi-tasking system. Staying within a single development environment our debug and analysis tools help to maximize developer productivity and reduce cost.
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 and she has developed presentations for the Embedded Systems Conference and ICSPAT. She has held a variety of production, technical marketing and writing positions within technology companies and agencies in the Northwest.