The value-add of commercial Linux distributions in embedded military applications
StoryApril 02, 2008
Vince Hauber
Concurrent Computer Corporation
Roll your own embedded Linux? It may not be a good idea if you value support, robust tools, and future enhancements. In long-term military programs employing technology upgrades, it's best to stick with a commercially supported distro.
As the popularity of the Linux operating system continues to
grow in embedded applications, developers have a choice of using a commercial
Linux distribution or a free kernel.org version with open embedded tools.
Surveys have shown that free public Linux deployment is expanding in embedded
apps. So what must the commercial distributions offer to keep and grow market
share? We will examine three key factors: support, tools, and feature
enhancements.
Figure 1
(Click graphic to zoom by 2.2x)
|
|
Support
Linux distributions that have a reputation for outstanding
support certainly have an advantage over companies with support groups that
behave like black holes. But is support from a distribution an important
discriminator in embedded projects?
Enterprise data centers using commercial Linux distributions
have a steady, ongoing need to integrate new platforms and manage evolving
application software services. They routinely encounter system problems that
justify the need for Red Hat and Novell support contracts. Many embedded
military projects, however, are more static and technically bounded. Once an
application has been proven, there may not be a lot of subsequent change that
requires long-term kernel support outside of the core developer group. On the
other hand, if a defense project will require integration of new COTS platforms
over time, then the support of a commercial distribution may be desirable to
provide maintainability and stability.
Recent VDC survey data indicates that the number of new
embedded projects using a Linux operating system is growing at about 50 percent
annually while non-Linux embedded operating system use (RTOS or EOS) is
relatively flat. A more striking statistic is that about 80 percent of the new
Linux-based projects are expected to use a free public Linux, as opposed to a
commercial distribution. This would imply that the need for formal support, from
a commercial distribution, for example, is often not a critical necessity in
project life cycle planning. Over time, more project engineers have become
kernel and driver savvy, allowing embedded projects to be supported from within
the providing organization. Required Linux drivers for processor and I/O
devices are routinely available from their hardware vendors.
There are two schools of thought on embedded solution support.
One view is that embedded developers want to work at the API level and leave
the kernel support to the distribution vendor. Another view is that
availability and examination of kernel source code are important not only to
fix problems, but also to fully understand kernel behavior in relation to oneís
application. There are really two different classes of problems – general
kernel problems and problems that are unique to an embedded platformís
hardware, drivers, and application software. Support for the latter class
usually requires the knowledge and expertise of the embedded developing group
itself.
Superior tools
If not support, then are tools the likely enduring value-add?
Embedded solution designers require a wide variety of tools for
cross-compiling, debugging, in-circuit emulation, memory allocation, flash
updating, and overall project management. Examples of currently available
commercial tool products are Wind Riverís Workbench Development Suite,
MontaVistaís Platform Developer Kit and Application Developer Kit, LynuxWorksí
Luminosity IDE, and the NightStar debugging and analysis tool kit for
Concurrentís RedHawk Linux. Alternatively, a large variety of free tools are
available from any number of embedded Linux websites.
To maintain and grow market share, commercial Linux
distributions must continue to invest in user-friendly, productivity-enhancing
Eclipse-based embedded toolchains that provide value over both free open source
and in-house developed tools.
The one-stop nature and the vendor support of the commercial
tool kits currently provide a value-add over free, open source tools. The good
news for commercial Linux vendors is that use of roll-your-own, in-house tools
in embedded projects is down, but the downside is that more developers are
using free, user-community, Eclipse-based, embedded tools. Commercial vendors
need to continue to enhance their Eclipse IDE-based toolchains to keep ahead of
the features of improved free open tools. One of the natural advantages they
can offer is a complete tool set that will be supported and compatible with
evolving releases of the Linux kernel.
Enhancements to standard linux
The feature advantages of traditional real-time operating
systems, both Linux and non-Linux, were lessened by the introduction of the 2.6
kernelís improved preemptability, scheduling, and synchronization features. But
the 2.6 kernel is still not a true RTOS. Several companies seek to provide
value-add by offering kernel enhancements to offer guaranteed real-time
performance for applications that consistently need response in the less then 20
microsecond range. For example, Wind River offers real-time in a hybrid
solution -- its Real-Time Core product (formerly RTLinux) running together with
standard Linux. Concurrentís RedHawk Linux offers a single-kernel approach that
achieves guaranteed response via enhanced kernel preemption and shielding
features. Real-time Linux companies also deliver enhanced versions of open
source I/O drivers to meet time-critical requirements.
Other value-add features offer opportunities for commercial
distributions to compete with free Linux in mission-critical applications. Features
like DO-178 safety certification, POSIX conformance, EAL Common Criteria, and
MILS security, although costly to implement, are not likely to be available in
any free Linux form.
Commercial Linux in the embedded military market
Commercial Linux distributions are seeking to grow revenues
in military embedded markets while, at the same time, developers are evaluating
the use of free and community-supported software. Linux distributions must find
creative ways to add value to counter the impact of ìfree.î We believe the
answer is enhanced tools sets and value-add kernel features. From these
offerings, the support revenue will follow.
Vince Hauber, a
senior product manager with Concurrent Computer Corporation, has more than 40
years of experience in system software and platform solutions. His current
responsibilities include Linux systems product management and technical sales
support. Vince has a BS degree in Mathematics from Loyola College in
Maryland. He can be reached at vince.hauber@ccur.com
Concurrent Computer Corporation
678-258-4000
www.ccur.com
