There are many reasons for the accelerating trend to move Residential Gateway (RG) products to Linux-based platforms. These RG products include traditional router/gateway, firewall, and low-end functions that were handled by a proprietary RTOS, as well as new devices needing Virtual Private Network (VPN) support, Voice-Over-'X', and other value-add features.

Built as a networking operating system from the ground-up, Linux caters to a range of communication products throughout the embedded sector. An open system compliant with international standards, Linux is ideally suited for retrofit to existing boxes, and as a scalable OS platform for the emergent networking products of tomorrow.

Background

Until recently, Residential Gateways (RG) and related class of networking products, were little more than glorified network hubs, or offshoots of larger network switching products. Today, these devices have matured to become a critical device for residential consumers, telecommuters, home office workers, and small- to medium-size businesses (SMB). As these products continue to evolve, concurrent efforts to promote broadband access via industry initiatives such as CableHome, and legislative mandate such as the 'Broadband Bill', exacerbate the need for capable, scalable devices that will fuel the next wave of data and entertainment delivery.

Historically, RG devices were built on a proprietary real-time operating system (RTOS), or equivalent 'home grown' code developed by in-house engineering teams. Such operating systems are hard-pressed to compete with an OS covered by an open source license, as it is difficult for their relatively small engineering team to support the variety and complexity of newly emerging features. The accelerating appearance of new products, services, and protocols such as IPv6, multicasting, and a desire for multiple-use services (data over cable, voice over IP, etc.), places great pressure on the ever more antiquated proprietary OS.

The proprietary RTOS vendors are finding it difficult to keep pace, due to their limited development communities and skyrocketing costs. In-house OS teams face the same issues. In fact, even if they could keep up, the 'opportunity cost' differential worsens. The end result is products with low differentiation with market delay to the product's introduction.

Even companies that try to use 'free' Linux, by using off-the-shelf versions and attempting to squeeze them into small devices face a number of daunting challenges. Although packaged with thousands of neat and useful software modules, desktop or server distributions of Linux are not suitable for embedded use. Modules that might be used for a 'desktop box' have not been downsized and optimized for an embedded device; such products must often run Linux in less then 2MB of space.

It is a fact that current RG products are commodity items, with narrow profits and little product differentiation. The saturation point, if not already here, will be attained only through further price erosion and market shakeouts. Many existing devices are inherently unable to address emerging markets. Small, low-cost devices often lack the computing power needed for VPNs with any reasonable level of security or encryption. The same is true for new multicasting protocols and multimedia devices. These devices are the next wave, as they open new markets and revenue streams.

Moreover, various hardware players have developed embedded networking-centric silicon. With faster processors bundled into system-on-chip forms, performance is rising, while costs are being driven to razor-thin margins. The proprietary RTOS faces even more challenges within this space. Foremost, due to the binary-only availability of their code, they are unable to react to the rapid pace of innovation demanded by Original Design Manufacturers (ODMs). Second, unlike Linux, proprietary RTOSs were often not originally built to accommodate the scalability that new devices demand.

An important attribute of the software platform is device scalability. Devices are often redesigned or repackaged to reduce costs or to add new features; such an approach can reduce design time and costs, but only if the repackaging can be done quickly, accurately, and with new hardware support. Few RTOSs or proprietary desktop OSs can do that.

An accessible development environment is also a must. When time-to-market was not as critical to success, there was little need to reduce the design, integrate, test, and redesign loop used in the past. Teams must be able to research, develop, and deploy, preferably on the same software base, in order to attain the most robust products and the shortest possible delivery time. Few RTOSs or proprietary OSs can do that.

In the fight to maintain or improve market share, one must reduce development costs, share design costs across multiple devices, and ensure suitability to task (the ability of the device or operating software to function with adequate performance headroom). At the same time, engineers must overcome issues such as loss of RTOS predictability and performance in network and disk operations, a common failing of most current-generation operating software.

A common surprise is that one might be tempted to think that an RTOS provides 'real-time' response. In fact most do not when network or disk operations are in use. As new devices add network, disk, and file systems, their footprint grows dramatically, and their real-time predictability is lost. Finally, unless devices have very careful design, existing RTOS systems use a memory structure that can result in system crashes and security risks.

In other words, for all of their expense and inflexibility, proprietary operating systems are unable to meet the demands of new devices, and simply don't offer any advantages. Clearly, an alternative is needed.

Emerging markets

In the past, 'footprint' — the amount of memory used by the software — was king. Although still important, other components of the network box are pushing footprint-only considerations aside. Next-generation RGs, and multimedia devices place heavy demands on vendors for new features, product differentiation, scalability, adherence to standards, time to market, security (against attack and theft-of-services), and of course, costs (including hidden costs of extended software development). New devices and services in Residential Gateway markets demand . . .

• Better router, firewall, and security interfaces

• Support for new protocols, services, and interoperability with standards

• Multicasting and other 'difficult' routing capabilities

• Support for VPN systems with strong encryption

• Enabling new services and associated revenue streams

• Enabling new paradigms such as the Virtual Office, and Telecommuting

• And all the while, creating such devices faster, better, and at lower cost

Linux is widely recognized as an operating system that extends across such an expansive market. It is mature, scalable, robust, and supports desktop-research efforts as well as deployment of devices from the smallest cell phone or home router, to large multiprocessor super-scalar and high-availability server clusters. No other operating system offers this breadth of capabilities.

Several companies now focus on delivery of embedded Linux. The companies that are successful in this space can deliver a complete solution. That is, one that includes the operating system, tools for specific boards (generally called Board Support Packages, or 'BSPs'), and a complete set of integratable components needed to rapidly and accurately construct new RG products.

The selection of components for dedicated embedded devices is very different from that used for the desktop. There are tighter code limits, and fewer opportunities to fix or upgrade the software. Engineers who are new to Linux often try to use a desktop distribution and 'packages', and squeeze them into a small footprint. The results tend to be poorly integrated and use more memory than optimized components, because there can be hidden dependencies that are not found by inexperienced engineers, and because the packages are inappropriate from the start. Add to that lost time-to-market, higher device costs for additional memory, and lack of differentiation because engineers working on 'reinventing' are not adding features. Development teams can become artificially inflated too.

Yet companies insist on continuing. Why? The simple answer is because of the overwhelming advantages of Linux itself.

Competitive options

It is evident that RTOS vendors can't keep up, desktop OS's are expensive and slow, and in-house 'reinvention' teams waste effort. RTOS and desktop vendors often do not adhere to standards, have poor ROI, and experience development difficulty and delays due to their necessarily smaller engineering teams. And yet, at the end, their substantially higher prices wipe out any other possible advantages.

Smart companies now have a choice: Linux provides the features needed for next-generation devices, today. Linux was designed for network operation. Adherence to standards is an overwhelming advantage compared to other OS's. Linux is already in wide use for many of the functions now moving to small devices. Linux also provides source code that allows companies to modify and tune the system when necessary, and to ensure security of products in ways that no proprietary vendor can match. Linux is the underlying 'system of choice' for many companies and their newest product offerings.

Yet embedding Linux can be difficult. Successful products need more than just an operating system: they need integrated solution components, and strong tools. Even a basic kernel configuration has more than 1,700 configurable options. Learning about those options and their interactions takes time. Basic RG components adds hundreds or thousands more options. Assuring that they work well together takes even longer.

Several companies offer various embedded solutions to assuage the aforementioned challenges. Solution stacks alone are just one variable to the equation. When the stack is combined with appropriate tools such as configurators and debugging support the result is a much faster and better product overall.

REDSonic offers such a combination in the optimized RG stack, SecureSOHO, and the companion embedded Linux development tool, RED-Builder. With some recently announced restructuring, Utah-based Lineo also offers a development tool, Embedix, and has announced that they will begin providing various vertical market software solutions — one of which provides stacks for RG products. Arcturus Networks, the core creator of the MMU-less uCLinux embedded distribution, also provides an RG stack, related firmware, and some product reference designs. And through their subscription model, MontaVista Software provides customers with the tool called TCT (Target Configuration Tool) and Linux support packages (LSPs), which provide core embedded Linux functionality for a variety of architectures and board platforms. Additionally, there are a few companies who do not specialize in Linux, but also offer products catering to the RG market for various OS platforms. For example, Israeli headquartered Jungo provides such an RG stack, but does not offer tools for embedded Linux development.

Companies in the RG space should ask the tough questions in order to select an appropriate software partner. It is not merely enough to have Linux booting to on the design platform. A designer will gain significant size and performance advantages for their product by seeking a partner who offers a complete combination of operating system, tools, and 'stack' solutions. Working closely with the ODM, the software partner can assist in the selection of components for routers, gateways, and firewalls, as well as accurately determine a growth path to new devices such as digital video boxes, multimedia equipment, and network appliances.

Conclusion

Linux is already penetrating markets that were, until just recently, the province of proprietary operating systems. It has done this because of exceptional stability, scalability, suitability to task, access to source code, and overall value. Yet some companies have been overwhelmed by Linux, due to those same characteristics, and the effort required to learn and optimize devices using Linux.

Embedded Linux providers can relieve those pressures, and provide Linux for new products more quickly and effectively, freeing development teams to focus on the addition of new features. Due to the scalability of Linux, those features are easier than ever to incorporate into new products, because they can be developed on widely available desktop systems, and then migrated to the target device.

Stack oriented Linux companies can provide the ODM with the optimized embedded Linux base, scalable communication stacks, and the tools to configure both into powerful new solutions.

Smart product design companies are increasingly recognizing the value this combination represents.

Copyright © 2002, REDSonic, Inc. All rights reserved. Published by LinuxDevices.com with permission.

---

About the authors: John Drabik is former VP and CTO for Digital Media at Lineo, and was Architect and Chief Engineer for their Residential Gateway and Digital Media plans. A strong believer in open standards and open source, balanced with the business development opportunities, John has led embedded development for several top tier companies including Texscan MSI, Evans & Sutherland, and EDO Western. John has written articles for LinuxDevices.com and other leading publications in the embedded space, and is a Consulting Engineer for REDSonic's RG Products Division.

Prior to joining REDSonic, Kevin Tobin developed hands-on appreciation for the networking value of UNIX-based platforms by spearheading the integration amongst dissimilar property management, PBX, and POS systems for several top-tier destination resorts in the hospitality industry. Tobin currently conducts business development and marketing activities for REDSonic's U.S. headquarters in Santa Ana, CA, and holds a B.A. in English Literature from the University of Arizona.

---

 
This article was originally published on LinuxDevices.com and has been donated to the open source community by QuinStreet Inc. Please visit LinuxToday.com for up-to-date news and articles about Linux and open source.