ELC: Trends in embedded Linux [LWN.net]

By Jake Edge
April 16, 2008

Henry Kingman, editor of LinuxDevices, opened the Embedded Linux Conference with a look at the trends in embedded development since he started covering the subject in 1999. Based largely on the annual surveys run by LinuxDevices, his keynote speech highlighted the growth of Linux as an embedded operating system as well as where it is headed in the next few years.

The conference, which started April 15 in Mountain View, California, gathers around 175 embedded developers for three days of talks on a wide variety of embedded topics. Sponsored by the Consumer Electronics Linux Forum (CELF), the conference has become the premier technical conference for the ever-growing embedded Linux community. Each day has a keynote, with kernel hacker Andrew Morton and CELF architecture group chair (and conference organizer) Tim Bird rounding those out, followed by a half-dozen presentations slots, with three parallel presentations.

Bird introduced Kingman as one of the main providers of news about embedded Linux, relating that LinuxDevices and LWN.net are his "two main sources of information" about the community. Bird marveled at the body of work that Kingman has amassed: "this guy is prolific". He also reminisced a bit about the early days of embedded Linux, starting with his days at Lineo to his current work at Sony:

It was hard to get people to pay attention to Linux, now Sony is putting Linux into almost everything.

Kingman acknowledged Bird's introduction, but said that he didn't know "if that makes me an expert in the forest, or lost in the trees". He looked back to a 1999 San Francisco Bay Linux Users Group meeting with Linus Torvalds as the featured speaker. Kingman said that Torvalds wanted Linux to be a desktop operating system but that he saw the embedded space as the big growth area.

Later that year, Kingman attended the first LinuxWorld conference where he saw some folks from Transmeta talking about squashfs and cramfs. An article he wrote about those filesystems was published by Rick Lehrbaum, founder of LinuxDevices. That was the first of more than 3000 articles Kingman has since written for LinuxDevices.

Kingman then presented the results of the most recent LinuxDevices reader survey. The survey gathers information about what LinuxDevices readers are doing or planning with regard to embedded Linux development. It has been run for eight years, providing some interesting information on changes in the readers' attitudes over the years.

Usage of Linux in embedded development projects crossed a threshold this year, with more than 50% of the 812 respondents saying that they are currently using it. Usage of Linux has been growing year over year, but didn't cross the halfway mark until 2008. More than 61% believed their company would be using Linux within the next two years.

The ARM family of processors has continued its growth with 30% of the readers using it, while 25% are using x86 variants. ARM overtook x86 three years ago; that trend looks to be continuing with respondents seeing 31% ARM versus 23% x86 over the next two years. Kingman said that he thinks Intel is trying to reverse that trend because spending on consumer devices is predicted to "outstrip IT spending".

There were a couple of questions asking where respondents obtain the version of Linux they use in their products. Ubuntu has a somewhat surprising share at 8%. For a relatively new distribution that is not specifically targeted at that market, it stands out, as does its predicted growth to 10% over the next two years. Kernel.org at 16% and Debian at 14% are the leading sources, with uClinux tied with Ubuntu and MontaVista and Fedora at 6% each.

Unsurprisingly, per-unit royalties were not popular with two-thirds of respondents being unwilling to pay those, but 60% were willing to pay for development and support of embedded Linux, so it is not just the free-beer aspect that is drawing companies to Linux. Most (45%) get their sources as a free download from a community site like kernel.org or handhelds.org, with 18% getting them bundled with their hardware. Only 11% said that cost was the greatest influence on their choice.

Legal threats are still on the minds of some, with copyright or patent concerns being considered a significant threat to roughly half of the respondents. SCO has fallen off the radar, with only 2.5% thinking that it is still a threat. "None of the above" was the big winner, presumably meaning that there are no significant threats, at 40%.

Kingman finished with a request of the embedded community to let him know what things should be covered in more depth and any additional areas they wish to see covered. He is looking for input on what the community wants to talk about: "we want to be your website."

(Log in to post comments)

ELC: Trends in embedded Linux

Posted Apr 18, 2008 20:55 UTC (Fri) by giraffedata (subscriber, #1954) [Link]

per-unit royalties were not popular with two-thirds of respondents being unwilling to pay those, but 60% were willing to pay for development and support of embedded Linux

Sounds like a dumb question to me. I'm sure it depends on the numbers. Respondents apparently imagine some particular numbers, but we don't know what they are and there's no reason to believe they're viable. I'm sure if the choice is between a 1% royalty and a $100K/yr contribution to development and support, smaller and more speculative companies would be great champions of per-unit royalties.

Only 11% said that cost was the greatest influence on their choice.

Again, I have no idea what question they were answering. How do you rank the influence of two different factors? And if you're a halfway decent business person, cost is the only factor. All those other factors are contributions to cost.

ELC: Trends in embedded Linux

Posted Apr 20, 2008 12:42 UTC (Sun) by man_ls (guest, #15091) [Link]

Ultimately everything can be evaluated as money, but cost should never be the only factor. Take legal risks: not all of them can be reduced to cost; the risk of jail time is a good example. The loss of reputation associated with a lawsuit can be more important than the immediate gains.

In the survey there are two factors above cost: "Availability of free, modifiable source code" and "Development tools". The first represents technological independence, which is hard to quantify as a cost but which can be an important strategical (long-term) factor. The second eases development and is probably a short-to-medium-term concern. If the "cost" factor is meant as the immediate payments due to licensing, then it makes a lot of sense.

ELC: Trends in embedded Linux

Posted Apr 20, 2008 23:16 UTC (Sun) by giraffedata (subscriber, #1954) [Link]

I take your point that businesses consider things other than cost, but lost reputation is a poor example, because in business, reputation is money. You spend money to protect your reputation when it would cost less than a damaged reputation would.

Also, I can believe that some respondents read "cost" as "licensing fee," though that's not the way skilled business persons use the term.

Nonetheless, the factors mentioned in the article all seem to be business factors to me. An embedded system maker would reduce them all to some common currency (not necessarily so far as to put numbers in a spreadsheet, but essentially the same), add them up, and take the configuration with the lowest sum.

So I still can't see how to rank the factors against each other. I think at a high enough licensing fee, every respondent would dump an option that has great development tools. So how can you say licensing fee is less important than development tools?

System factors

Posted Apr 21, 2008 6:18 UTC (Mon) by man_ls (guest, #15091) [Link]

I don't know how other companies do it; at my current job they are very brand-conscious, and will go to great lengths to make sure customers are satisfied. Even when it costs the company some money (and against an unquantified loss of reputation). Maybe that is not the best way to do it, but it works for us.

You are right that generic factors cannot be ranked against each other. In this case I believe that factors are to be taken more matter-of-factly: when an embedded developer chooses one OS from those currently in the market, what is more important to them? It seems that in the current crop license fees are not such a differentiator, while development tools are.

Given that some options (such as Debian) are free of licensing costs I take it that a lot of people would be willing to pay for good dev tools as long as code is free. This is only my interpretation; you are right that it can be confusing as to how each factor stacks against the rest.

ELC: Trends in embedded Linux

Posted Apr 24, 2008 5:27 UTC (Thu) by joern (subscriber, #22392) [Link]

While many things can be reduced to cost, cost is usually trumped by at least two other
factors: loss of job and loss of company.

Each decision has to be made by some individual person.  Remember the "noone got fired for
buying IBM" slogan?  It simply reflects that people don't mind costing the company more if it
saves themselves.

Loss of company trumps even that.  Bad business decisions can be camouflaged, some people have
a personal charisma that allows them to outlive many bad decisions in a company.  But no job
can outlive the company.  So when given a choice of saving the company $$$ with 99% success
rate, while the remaining 1% will break the company's neck, don't expect the savings to be
taken.

One exception is when the company is about to go bankrupt and not saving $$$ would kill the
company anyway.