RTC Magazine

RTC: LynuxWorks is well known as a company that long had a successful real-time operating system and has incorporated Linux into its technology portfolio. Can you describe how Linux fits with your real-time offering LynxOS and what issues the two technologies address both separately and as complementary products?

Singh: LynuxWorks is unique among the embedded RTOS vendors in the way we embraced Linux when it appeared in the embedded industry in 2000 and developed a strategy of providing consistent embedded solutions incorporating Linux and our RTOS technology. Other vendors have subsequently developed similar dual OS strategies, but none of them provide the architectural elegance of our technology portfolio.

LynxOS, our flagship product, is a hard real-time POSIX-conformant operating system targeted at embedded applications, while Linux was designed as a general-purpose operating system targeted at desktops and servers, although it has been adapted for the embedded applications, including our own BlueCat Linux product. However, both LynxOS and Linux were developed from the ground up to support the Unix software interfaces and the Unix look and feel, that is, as Unix work-alike products. Thus, both are highly compatible and interoperable, and LynxOS even supports the Linux ABI (applications binary interface) so that shrink-wrap Linux software can run unmodified on LynxOS. LynxOS-178, which targets safety -critical applications, will also support the Linux ABI in the near future.

Inder Singh, CEO of LynuxWorks

In addition, BlueCat Linux, LynxOS and LynxOS-178 are all supported by a common Eclipse-based development environment with a powerful suite of tools. This provides developers with a consistent suite of operating systems that maximizes flexibility in architecting embedded solutions over a wide range of applications.

Many of our customers have taken advantage of this OS suite to build distributed systems that utilize both OSs. In some cases, BlueCat Linux is used on processors that implement the man/machine interface, while LynxOS is utilized with processors implementing the main real-time application. In others, LynxOS is used in the central performance-critical module, while BlueCat is used in the larger number of cost-sensitive nodes. Yet another approach for networking applications is the use of BlueCat on the control processor and LynxOS in the line cards. The developers are able to use a single common development environment, deal with a single vendor, and have the flexibility to partition the software to run a given module on either operating system during the development cycle.

The LynuxWorks product portfolio also provides an attractive migration path and reduces risk. Customers can develop their application on BlueCat Linux and migrate to LynxOS if real-time performance becomes an issue as system requirements or the workload change during a product’s life cycle.

RTC: Several years ago, there was an initiative by the Embedded Linux Consortium to define a standard embedded Linux platform. We haven’t heard much about that recently. Can you tell us if that effort is ongoing and if so, what the status is?

Singh: The ELC Working Group did a very nice job in coming up with the ELCPS specification for a common set of APIs that must be available to applications software on embedded Linux systems. The specification has been well received and there is strong interest in using this as conformant products become available. For an OS specification like this to become viable, you need three things: test suites to verify compliance, availability of conforming OS products, and conforming applications or users who demand conformance to the specification. The ELCPS has not yet succeeded in breaking this chicken and egg cycle. However, there is a widely felt need for this kind of a platform standard, particularly by vendors of Linux-based software products. The ELC is working on ways to move this process forward.

RTC: The open source movement (for lack of a better term) appears to have created something of a bandwagon effect. Now, in addition to the Linux community, we see players like Sun Microsystems and to some extent even Microsoft at least acknowledging its importance even though there seem to be widely differing ideas about what “open source” really means or should represent. Could you give us your views on what this phenomena actually means, in concrete terms to OEMs who want to get to market with reliable code and also control costs and time-to-market?

Singh: Open source software and Linux, in particular, have clearly developed unstoppable momentum in recent years. The low cost of such software (not quite “free” since there can be significant indirect costs other than license fees) is clearly a major attraction especially in the emerging world and for high volume consumer products. However, it is the free access to the source and the creation of a development community around software products of this kind that are the major driving forces. Open source projects have been most successful for foundational kinds of software such as operating systems, development tools and network protocols, which are widely used by developers themselves to build other systems, and where the “network effect ” is the strongest, i.e., the more widely a product is used, the more valuable it becomes. For an OEM who wants to bring an embedded product to market, it is important to take a realistic look at the real cost and schedule risk of using an open source product. Support by a well-established vendor who has a successful track record of supporting OEM customers is something to look for. The quality and reliability of the open source product need to be evaluated, as well as the value of access to the source for providing special functionality or for handling ongoing support of the final product. The availability of third-party software and a supporting developer community are significant benefits to evaluate as well.

RTC: With the proliferation of networked systems and devices, the importance of what we call “middleware” appears to be increasing. CORBA and OSGI come to mind as examples. Can you comment about the role of middleware for small networked devices and what kinds of functionality it will be expected to provide?

Singh: With the growth of networked systems with increasing functionality and complexity, the role of middleware, the common layers of software between the underlying operating systems in the individual devices and the various applications, is growing in importance. At the same time, this whole area is in a state of flux with many different approaches and emerging standards. Some of the driving forces are the need for interoperability, and the growing awareness that the complexity of these systems dictates the reuse of standard building blocks instead of trying to build each system from scratch in an ad hoc fashion. The middleware is intended to provide a common infrastructure or “plumbing”, to provide the common “glue” for connecting different entities such as programs and data structures, agents and entities, and so on. We see both CORBA and DDS (Distributed Data Systems—a recent publish and subscribe system standard from the Object Management Group) being used in many of the new defense systems with a strong Network Centric Warfare thrust. We also see growing interest in the OSGI framework, which provides powerful life cycle support for running multiple Java applications in networked systems and managing them remotely over the network, in application areas ranging from network-centric defense systems to telematics and network services targeting smart phones.

RTC: Perhaps a more immediate phenomenon is the gradual but steady move toward VoIP displacing the traditional POTS and/or circuit-switched networks. How important is the proliferation of VoIP for voice services to the eventual convergence of many additional different digital services, including video-on-demand and the various kinds of data services, all over the worldwide network?

Singh: VoIP is a first step toward an “IP everywhere” future for communications that takes us to a world where the distinction between data, voice and video becomes merely an issue of encoding. We have been sending data over voice channels using modems for a long time. Now voice and video are simply types of information that can be transferred over data networks instead, and can be shared and processed in myriad ways. This will unleash a smorgasbord of new products and networked services, many of which are hard to even visualize today.

RTC: One very big aspect of all this networking is, of course, wireless connectivity. It’s been said that, increasingly, the cell phone can be expected to be the platform for the delivery of various kinds of services. Several things: is a small handheld phone physically adequate to provide these services and what kinds of services will they be? What kind of bandwidth will we need in order to make these services attractive and what will be the “lingua franca” so to speak—the overarching software environment within which this will take place.

Singh: In the cell phone you have an intelligent device with a network connection and a user interface that is carried by more people than anything else on the horizon. This makes it the prime candidate for the next computing “platform” after the PC, where the long heralded convergence of computing, communications and entertainment can happen. Even with a limited screen size and keyboard, we already see the cell phone taking on the functionality of the PDA, e-mail terminal, camera, MP3 player and camcorde,r and even video games. I believe this is just the beginning; the next major step is a growing base of network delivered services including financial services of many kinds: the cell phone serves as the user device for accessing these services. Network bandwidth is a key limitation today. We will see continuing demand for higher bandwidth, at least up to several megabits per second to support video. There are several software environments that are competing for dominance in this explosive market segment including Microsoft CE, Symbian and Linux. I believe Linux along with Java will emerge as the dominant multi-vendor standard platform for this explosive market.RTC: Any discussion of networking, particularly wireless networking, brings up huge issues of security. If people really are going to be making significant financial and other transactions via mobile units like cell phones, where will security measures like authentication and encryption/decryption be implemented. In short, what kind of security infrastructure do you see as being needed to enable all these services we’ve been hearing about?Singh: Security will clearly be a critical issue as cell phones handle more critical network-based services including many kinds of financial transactions. The network security issue is inherently not that much harder than for the wired Internet, which is really pretty wide open. The main additional wrinkle is authentication since cell phones can easily be lost or stolen. I believe we will see an increasing use of biometrics to handle the issue. One clear advantage is that you can start with more reliable operating systems than Windows on the PC, which provides a very insecure foundation for PC-based systems.

RTC: The government obviously has some levels of security that it will not be willing to have used by the public, while it may actually mandate others as in the case of FIPS. What role do you think government security standards such as FIPS, MILS and others will play in the security of overall civilian systems?

Singh: The government mandates very stringent security requirements for handling classified data, particularly MLS (multi-level secure) systems that handle data at different classification levels such as secret and top secret. New technologies such as MILS architecture products, many of which are COTS (commercial off –the-shelf) products, are being developed for high robustness secure systems designed to handle this kind of MLS data. We believe some of these technologies will be found to be very useful for civilian systems that handle critical financial transactions as well as critical infrastructure such as power plants and distribution networks.

RTC: Where do you think we currently stand in terms of network security—not only in terms of embedded, connected devices, but also with the large host and network operating systems like Windows, Linux and Unix, which are an integral part of the overall networked world? How big an impediment is the current state of affairs to rapid development?

Singh: For all the talk and growing angst about security, I believe it is still not being addressed adequately. Perhaps it will take an act of cyber-terrorism or other disaster for this problem to get the attention and the investment that is required. The proliferation of Internet-connected embedded devices in particular has created many opportunities for malicious users to exploit security weaknesses in embedded software to gain access to sensitive systems. Such an intruder can access sensitive information, bring down a critical system, or gain control and modify its behavior in dangerous ways. Even embedded devices not directly on the Internet may be accessible through dial-up ports, or be connected to local private networks, which can become accessible over the Internet through PCs on these networks that may be connected to the Internet by individual users. Such embedded devices can be found in many critical areas: from the power grid and the communications infrastructure to power utilities, railroads, as well as chemical and nuclear plants. They include SCADA (supervisory control and data acquisition) systems, PLCs (programmable logic controllers), digital controllers, communications switches and intelligent devices of many kinds.

Ubiquitous network connectivity is what creates the growing challenges, but the greatest vulnerability is in the operating systems of the network-connected devices, along with the middleware and application stacks. The network itself can largely be protected through encryption. Operating systems such as Windows and even Linux, on the other hand, have been growing in complexity making them very hard to analyze and ensure the absence of vulnerabilities. Current system architectures also lead to large amounts of software running unprotected in system mode with full access to all the system resources. This includes device drivers and network software, often provided by many different organizations. The emerging MILS architecture based on a small separation kernel at the base of the system, which is small enough for formal verification, offers an elegant way to address this challenge. This is currently targeted at the most critical defense systems, but I expect this approach to find use in many mission-critical commercial systems over time as well.