Jim Ready, founder and CEO of MontaVista Software, is renowned as a pioneer in the real-time and embedded operating systems market. He co-founded Ready Systems in 1980, the company that developed VRTX, the embedded computer industry's first popular commercial real-time kernel. In this interview, LinuxDevices.com founder Rick Lehrbaum asks Jim about his views on the embedded Linux market in general and… about MontaVista Software's plans and strategies.

RL: You've said that MontaVista is going after the embedded Linux market. What application areas are you targeting?

Ready: A broad spectrum of embedded applications. And embedded, by our definition, is not the pseudo-embedded 'palmtop' space targeted by Microsoft with WinCE.

RL: OK, so what's your definition of 'embedded'?

Ready: Let's say 'that which differs in form and function from a desktop PC and its direct descendents and their use for general purpose productivity applications, as opposed to dedicated and/or control applications.' In other words, not the desktop, not laptop or palmtop, not servers, but all kinds of specialized devices that differ pretty significantly from what you do in normal desktop or server computing. These tend to be specialized, dedicated applications, often based on specialized form-factors or custom hardware. For example, a server based on CompactPCI (CPCI) boards and offering hot-swap high-availability is a much more specialized system than a rack full of motherboards boards running a web-server application.

RL: How do you plan to be successful in a market based on royalty free, open-source software?

Ready: Our strategy is simple: we will provide 100% pure Linux, and without runtime royalties. You might say, our motto is to 'learn to live in the Linux world _and_ prosper'. We think it's very possible to have a healthy business model based on Linux without resorting to a licensing approach, contrary to what some companies are doing. Basically, we don't want to kill the goose that laid the golden egg.

RL: … and that golden egg is …

Ready: … open source Linux, and the open source developer community. And all the benefits that derive from open source software, from the customer's perspective. Looking back, I was associated with inventing the 'traditional' real-time operating system (RTOS) software market, which is based on a proprietary/royalty model. At MontaVista, we feel very strongly that it's essential to be in harmony with, and to support, the open-source nature of Linux.

RL: Could you be a bit more precise about MontaVista's business model? Given that you're not going to charge royalties, how will you generate revenue?

Ready: The opportunity with Linux is to not increase the cost of the customer's product, but rather to decrease it, by not having runtime royalties. Our goal is to develop a business relationship with these customers that satisfies our financial needs in other ways — via our 'subscription' program. We really don't see any restriction on MontaVista's ability to generate revenue by not going down the royalty path. It's really simple: customers just don't want to add cost to their products — they want to take cost out of their products! Why not take advantage of the established Linux model (open source, no royalties) and do business in a way that's more compatible with customer requirements?

RL: From your experience (at Ready Systems and elsewhere), what percent of an embedded and real-time operating system company's revenues typically come from royalties, vs. from support and services?

Ready: It's typically a lot less than one might expect: around 20%. Thinking about a Linux based company like MontaVista, this might imply a 20% reduction in sales as compared with a royalty-based OS company. But of course, in the case of Linux, there's a much broader appeal. We also expect renewals of our subscriptions to generate more revenue on an ongoing basis, and at a steadier rate, than what the traditional RTOS companies have derived from their heavily front-loaded sales of binary licenses and SDKs. A really big problem for the customers (with the traditional RTOS sales model), is that all the money's paid to the supplier up front — and that makes it hard as hell to get good support further down the road, when there's no more money changing hands. We think our subscription model is vastly superior to the traditional RTOS company business model — for both us, and our customers. It's just a better way to do business.

RL: What do you see as MontaVista's primary value add?

Ready: The #1 issue to companies trying to deploy embedded product, these days, is time-to-market. With our subscription model, we become a partner with the customer. We provide them a distribution, regular updates, and tools support. All this helps the customer to reduce time-to-market. And that's what the customer's willing to pay for — faster time-to-market. In short, we're not increasing the customer's product costs, and more importantly, we're helping the customer hit critical market windows.

RL: What's the status of Hard Hat Linux (MontaVista's Linux distribution)?

Ready: Hard Hat Linux officially released and there are actually companies already using it in products. At this point we've got two main distributions, both of which are available via MontaVista's subscription programs. One, for 'x86', is qualified on selected board-level products. This includes Force, Ziatech, and Motorola Computer Group in the CompactPCI space, and WinSystems and Ampro for PC/104.

RL: And the other distribution?

Ready: The other distribution is for PowerPC. It supports PowerPC reference platforms and off-the-shelf CompactPCI products, but tends to be most popular with Motorola and IBM high-integration chips, like the PowerQUICC.

RL: Could you clarify what do you meant when you said your distributions are available via subscription?

Ready: When you subscribe to Hard Hat Linux from MontaVista, you get the selected distribution plus all its new releases (and bug fixes) for a year. Also, depending on the level of your subscription, you get varying degrees of responsiveness from the MontaVista support organization.

RL: Could you be a bit more specific about the 'varying degrees of responsiveness'?

Ready: We offer three different levels, differentiated by how fast we respond, how proactive we are with updates and bug fixes, etc. So the customer only needs to pay for the responsiveness they want or need.

RL: How much does it cost?

Ready: The program is intended for design teams in major development companies that want a top tier relationship with MontaVista and are willing to pay for that. The base level is $5,000 per year, per developer. We have several customers who've purchased packages of twenty-five or more subscriptions, for their product teams.

RL: Could you tell us about some of your design wins?

Ready: Our areas of greatest success have been in three markets. The first is the telecom/internet infrastructure equipment market, largely based on CompactPCI hardware. The second is that of internet appliances, using custom hardware. And the third is in the market we call 'A-to-Z embedded' — which includes instrumentation and control systems, like mass spectrometers, PCB inspection systems, and the like. Did you hear about the Kerbango Systems' new 'Internet radio'? It's a great example of a post-PC internet appliance with 'Hard Hat Linux Inside'.

RL: Is there anything different about these embedded Linux applications, relative to what you saw previously in your VRTX experience?

Ready: Yes. Many of the applications we see for embedded Linux are leveraging internet technology and a web based, rather than a serial or dedicated graphics hardware (e.g. VGA) user interface. It's very clear that the extensive networking and IP capability of Linux are going to be very powerful factors in its success in the embedded market.

RL: Has MontaVista developed a lot of original Linux software, or are you primarily distributing a standard Red Hat (or other) distribution coupled with your subscription support and services?

Ready: Hard Hat Linux contains a significant amount of support for booting Linux out of flash, including modifications for flash file systems, scripts and documentation for compressing the file system, and so forth. Also, there's Hard Hat Net, which provides an extremely valuable inter-processor communication technology for CompactPCI systems (which, by the way, we've released under the GPL open-source model). The differentiation between Hard Hat Linux and 'standard Linux' will become increasing pronounced over the next few months. I might mention that in the last month alone, we've added 20 software engineers who are busily engaged in this process.

RL: What are some of your plans for future enhancements to Hard Hat Linux?

Ready: Some of the things we're working on are improving the real-time performance of Linux, reducing its foot print — we're already fitting in less than a megabyte of run-time footprint — and further support for CompactPCI (including high availability and special services). We're also engaged in porting Linux to non-Intel microprocessors like Motorola's 405, 823, 860, and 8260 CPUs, for deeply embedded custom designs. And we're working on tools that will help customers quickly and easily customize and deploy their Linux applications.

RL: How much of what you're developing at MontaVista will you release as GPL open-source software?

Ready: All of it! Read my lips: 'We believe in this model!' We have the strong conviction that if we make Linux altogether better, it benefits the whole market — which grows the pie for everyone. I think companies that try to play both sides of this issue will fall by the wayside. The customers recognize that this is the right thing to do. Why in the age of Linux would customers allow themselves to get locked into proprietary software that they can't control? The customers are trying to get away from these proprietary locks. As for royalties, it's just not right to try to charge royalties for a Linux distribution.

RL: What's your projection of the size of the embedded and real-time operating systems market?

Ready: The measured market for out-sourced embedded and real-time OSes, this year, is said to be around $1B. Add to that at least another $1B for the costs associated with home grown (in-house developed) OSes. So, its you end up with a market that's at least several billion dollars.

RL: … and growing fast?

Ready: The growth rate of Linux is phenomenal. A recent IDC study says that between 1998 and 1999, Linux grew 190% in the server and desktop markets — and that was five times the growth rate for Microsoft NT. That's staggering, when you consider how entrenched Microsoft is in those markets. Imagine what the growth rate must be for Linux in the embedded market, where Microsoft isn't a dominant player — in fact, where there's never been a single dominant operating system? In thinking about this, bear in mind that the embedded market shows very little 'loyalty' to any of the proprietary players that are present there. I expect to see Linux growth on the order of quadrupling this year, in the embedded space. By the way, don't expect to be able to measure this success right away, since in the embedded market there's a lag of 12-18 months between design-in and volume shipments. But the design wins will show it, if you could come up with a way to count them.

RL: As you invest in adding value to your Hard Hat Linux distribution, will you begin to diverge from mainstream Linux and therefore stand the risk of losing out on being able to leverage all the advantages of 'standard Linux'?

Ready: It would be dumb for us to do anything that would sacrifice any of the benefits of the core Linux kernel, libraries, and utilities. We are basically shipping mainstream Linux, built from the standard Linux kernel sources. It boils down to the notion of what a Linux distribution is. Distributions of Linux are like flavors. There are desktop flavors, server flavors, firewall flavors, and so on. What we've done with Hard Hat Linux, is create an embedded flavor. The thing that makes a Linux distribution 'Linux' is the kernel, the libraries, and the implemented API. That core is what we're talking about retaining. We are going to make Hard Hat Linux reflect its application base, with components targeted to embedded development in the markets we serve. While doing this, we will remain compatible with all the standard Linux tools and APIs.

RL: What do you think about Red Hat's EL/IX API, which is supposed to be capable of supporting Linux and eCos kernel (the Cygnus kernel for deeply embedded applications) within a single common API?

Ready: I believe in Linux. Real, 100%, pure Linux. Pushed as far as it can go. I've been extremely impressed with how flexible, configurable, and scalable Linux is. And don't forget, I'm the 'VRTX guy'. Embedded system resources are much richer now, than in the past. You can hardly find a new embedded system design that has less than 4 megabytes of RAM. This has really changed the concept of what is required for 'deeply embedded' applications, from a resource perspective. We don't need another operating system (i.e. eCos) and we don't need to burden Linux with an API that provides eCos compatibility.

RL: How does Hard Hat Linux support the problem of sizing Linux to match the varying needs of specific embedded systems?

Ready: The standard Linux kernel inherently provides sizing capabilities, thanks to the way Linus designed it. The filesystem is a different story. On a standard Red Hat distribution, for example, the filesystem may require 15 to 20 megabytes of space. Not that it's not a mistake; it's just that they've set it up for general purpose desktop or server environments. You can rebuild the library to leave out what's not needed (such as debug information). If you do, you'll end up with a much smaller system requiring only a handful of megabytes of memory. Doing this requires reconfiguring the libraries, and you'll want a process to manage the configuration on an ongoing basis. Currently, we offer a choice of several 'canned' options for various types and sizes of systems. Over time, MontaVista will offer tools that automate the process of generating a broad range of embedded system configurations.

RL: Any final words of wisdom?

Ready: Don't kill the goose that lays the golden eggs. Don't contaminate Linux. 100% pure Linux. No royalties.

RL: Thank you!

 
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.