SMALL FORM FACTOR FORUM
Where Have All the RTOSs Gone?
COLIN MCCRACKEN & PAUL ROSENFELD
Page 1 of 1
Gone to graveyards, every one? Or so it seems in the 32-bit embedded space these days. A real-time operating system, unlike a typical desktop operating system, is designed from the ground up around a real-time scheduler. OSs can be categorized by size, features, determinism and number of threads. Why should we meditate about RTOSs long-time passing? Because software requirements ultimately steer processor decisions and determine memory footprint.
Gone to Mentor?
A long time ago in a galaxy far, far away, VRTX was the ultimate versatile real-time executive from Hunter & Ready (aka Ready Systems), which merged with Microtec Research and ultimately got swallowed by Mentor Graphics. From the 16-bit 68K/x86 architecture wars to early MIPS-based set-top boxes to early Motorola ARM7-SoC-based cell phones to the Hubble space telescope, VRTX covered the entire span from microcontroller version to full-featured RTOS. With such an impressive slate of embedded apps, where did this beloved RTOS go?
Mentor, fundamentally a respected EDA/simulation and software development tools company, never managed to take VRTX to the next level. After burial, Mentor got a second chance with the acquisition of Accelerated Technology (the “other” ATI) and their Nucleus OS just as the dot-com bubble was bursting. Nucleus clings to life these days in embedded networking stacks, dealt the same Linux blow as the other RTOSs.
Gone to Intel?
Wind River’s VxWorks, dating three decades back to those same early days, got a major leg up during the PowerPC networking telecom/datacom boom of the late ’90s until the bubble burst. In its own PAC-MAN style, Wind River gobbled up major competitor Integrated Systems with pSOS and TakeFive Software, and finally got swallowed by a bigger gobbler. Along the way, Wind River adopted its own Linux framework, since you have to join ’em if you can’t lick ’em.
Gone to planes, trains and automobiles?
QNX, once a mainstay in medical devices due to kernel-level modularity, early adoption of POSIX APIs and its GUI, swept into Harmon’s car infotainment strategy, then spun out again. A number of telematics and navigation systems continue to use it. New owner Research In Motion (RIM) has big plans—the ultimate “smart car” wireless and infotainment platform. Meanwhile, Green Hills Software differentiated itself as a premier development tools vendor, then developing Integrity as a focused offering to a high-value segment—Avionics DO-178 certification. LynxOS migrated from HP printers to DO-178 as well, with a side dish of Linux.
Gone to data centers, board vendors, and pseudo- and non-real-time?
At its peak, RadiSys acquired Microware and its OS-9 operating system. High-availability real-time continues to survive as a niche. OS-9 seems to have a better position in its sandbox than the mother ship in this new world order of commodity express. The data center market appears to have swallowed the remains of VenturCom’s RTX extensions for WinNT and WinXP on the strength of Citrix Systems. Cloud computing, 64-bit virtualization to share the cost of expensive processors, hypervisors and software-as-a-service seem to be the order of the day, not real-time computing. “Real what?” you say?
Gigahertz computing reduces latencies and cache miss penalties. So with billions of transistors at our disposal, why not throw the glut of performance at a desktop-class OS and simply tolerate the occasional missed deadline from the long tail of the non-deterministic bell curve? Engineering students and recent grads in Silicon Valley are enamored by the latest venture capital bubbles—free Android apps and social networking. Alas, the science of real-time design has succumbed to realities of engineering budgets, time-to-market pressures, and a new generation of Visual C++ developers with affordable low-end and mid-range x86 processors that can actually meet and exceed 5- to 7-year lifecycle requirements.
Gone to the open-source community?
From real-time Linux flavors to Symbian to dozens of other open-source GPL purpose-built OSs, there is no questioning the impact of the open-source community over the years. Even Microsoft had to reduce WinCE royalties and open-source some modules. At the end of the day, it appears that the new generation of developers likes cheap, rich, open-source application platforms like Android and Linux more than hard real time. Or perhaps the real cause of death was the expensive royalties that came along for the ride? If the question “do you need real time?” is answered with “what’s the cost?” it’s clear that hard real time is not a hard requirement. Just don’t take the cover off the medical robot that is going to operate on you.
When will they ever learn?