Linux has rapidly come to be used in a wide range of applications in the industrial field, from domestic appliances to devices used in social infrastructure. However, the road leading here has not always been a smooth one. For many years, Toshiba has taken on difficult challenges related to power-saving, reducing startup times, real-time control, and more. We have contributed to the evolution of Linux by achieving these goals while creating new value. New challenges have arisen, such as long-term maintenance, improved reliability, and cyber-resilience, and our engineers are constantly pursuing further innovation. In this four-part running feature, we will look at the numerous difficulties Toshiba has faced and the measures we have used to overcome them.
In Part 1, we will focus on the challenges involved in using Linux in various devices and the history of our efforts to tackle these challenges.

Linux is an open source operating system (OS) developed in 1991. It has been refined by software engineers around the world, becoming widely used as an OS for servers and computers. The embedded systems used in domestic appliances and social infrastructure devices, on the other hand, have primarily used real-time operating systems (RTOS). Starting in the late 1990s, the industrial world also began to turn its attention towards Linux, which was praised for its openness, flexibility, and expandability. Linux intended for use in the industrial field which has been customized based on the specific functional requirements and hardware environments of overall systems is known as "embedded Linux." Unlike the Linux used on ordinary servers and computers, embedded Linux must also offer the following features.

1. Operation in environments with limited resources: Embedded Linux must operate in environments with limited CPU power, memory, and storage
2. Low power consumption: Embedded Linux must consume minimal power in products for which battery operation times or standby times are important
3. Fast startup: Systems must start quickly when devices are powered up
4. Real-time performance: Processing must be reliably completed within specified periods of time

From the late 1990s, embedded Linux started to be used primarily in networking devices such as routers, digital televisions, and mobile phones. However, initially, it was incapable of fully leveraging the capabilities of these devices. In 2003, Toshiba joined the CE Linux Forum (in 2010, this merged with The Linux Foundation, and until 2023 it operated as the Core Embedded Linux Project). It made major contributions to the capabilities of embedded Linux through its community coordination and its leadership, helping drive Linux's evolution.

Let's look at some of the typical challenges faced when using embedded Linux, the actual challenges that Toshiba has faced, and the measures that Toshiba has used to address them.

As mentioned earlier, some of the challenges in deploying embedded Linux are how to ensure that it can operate with limited resources, consume little power, start up quickly, and offer real-time performance. Furthermore, each of these must be achieved in a way that is matched to the actual device in which embedded Linux is used and the applications the device will be used for.

In other words, there is a different solution for each and every product. This presents embedded Linux developers with a great deal of difficulty, but it also makes the development process an interesting one. Let's touch briefly on the challenges involved in meeting these requirements and look at what kinds of measures Toshiba has implemented.

■ The challenges of power conservation and the measures used to achieve greater power savings

For embedded systems, power conservation is an essential requirement. For devices with constant power connections, in particular, a great deal of importance is placed on minimizing standby power consumption. For example, with a digital television, the amount of time it takes to watch a television program after plugging in the television and turning on the power differs from the amount of time it takes when the television is already plugged in and it is turned on using the remote control.

This is because when a television is turned on and its functions become usable, and then the television is turned off using a remote control, the television goes into standby mode. The next time the television is turned on, it recovers from standby mode (Fig. 1). Standby mode cuts off power to various television components to prevent excess power usage, reducing power consumption. Keeping the television in this mode not only conserves power, but also shortens the time it takes for the television to resume operation when it is turned back on.

Initially, it was extremely difficult to apply embedded Linux to systems with power-saving functions like these, because it lacked mature power-saving technologies. For example, at random times, there was a one-in-several-thousand likelihood that the television would be unable to restart from standby mode. Fixing this problem took quite a long time. From a user's perspective, of course if you push the power button the television should turn on. To make sure this happened, engineers performed round after round of improvements and confirmation, and now televisions can be reliably turned on and off while consuming minimal power.

■ Shortening startup times and developing methods for speeding up operation based on various requirements

Another important challenge was shortening the time between when power is supplied to a device and when the device actually becomes usable. When embedded Linux first started being used in devices, there was no tuning process. Instead, embedded Linux was applied as-is, so it was common for device startup times to range from dozens of seconds to several minutes.

Startup time requirements vary by product. For some products, the faster the startup time, the better. For other products, what is essential is that the product starts up reliably within a specified period of time. Startup time reductions need to be based on the requirements of each individual product. Toshiba developed various methods for shortening startup times to meet device requirements. These include the following.

• Reducing the amount of data that is read upon startup
• Reviewing the OS initialization process and enable parallel startup
• Creating startup state snapshots (hibernation images) and reading them (snapshot booting)
• Making only the minimum required functions available first, and gradually enable other functions in the background

When applying embedded Linux to products, we used different methods such as these depending on specific hardware and device applications to achieve appropriate startup times for individual products. In some cases, we have also further accelerated startup times that were already accelerated through the use of snapshots (Fig. 2). By switching to booting from a snapshot, we shortened the startup time from its previous 28 seconds to just 11 seconds. We then optimized the processing performed when booting from the snapshot, skipping unnecessary processes. This further shortened the startup time by 3 seconds, bringing it under the 10 second mark.

From around 2010, Toshiba began using embedded Linux in devices used in social infrastructure. One of Toshiba's social infrastructure-related products that used embedded Linux was an automatic train ticket gate^*1. Embedded Linux has gone on to be used in devices such as the industrial controllers used in transportation and power applications^*2. Social infrastructure supports our ability to live stable, secure lives. Due to its importance, the devices used in this social infrastructure have more rigorous requirements for performance, reliability, and long-term, interruption-free usage.

* 1: Ref. Toshiba Review "EG-5000 Automatic Ticket Gate with High Reliability and Scalability"
https://www.global.toshiba/content/dam/toshiba/migration/corp/techReviewAssets/tech/review/2010/10/65_10pdf/f05.pdf (PDF)(422KB) (in Japanese)

* 2: Ref. Toshiba Review "TOSMAP-DSTM/LX Next-Generation Controller for Thermal Power Plants to Achieve Efficient Operation and Low Environmental Burden"
https://www.global.toshiba/content/dam/toshiba/migration/corp/techReviewAssets/tech/review/2013/11/68_11pdf/a08.pdf (PDF)(312KB) (in Japanese)

■ Taking on the challenges of real-time control

Control processing that guarantees that a process will be completed within a defined period of time is called "real-time control." The performance needed to achieve this is called "real-time performance." When a device is described as having "high real-time performance," people often interpret this to mean that processing is performed instantaneously, but it would be more accurate to say that the device has "the performance required to complete processes within a stipulated amount of time." This time will vary from product to product. Real-time control is a critical requirement of control systems such as those used in industrial equipment. For example, imagine someone walking through the automatic ticket gate in a train station. In the time it takes for the user to tap their IC card on the gate sensor and pass through the gate, the gate must calculate their fare, determine if the user is allowed to pass through, and open for them. If it does not handle all of this in time, it can stall the user at the gate and disturb the flow of traffic.

In other words, each product must be designed to meet differing requirements for finishing certain processes by a certain point in time. Toshiba has drawn out Linux's full real-time performance potential and contributed to each product's ability to reach its target performance level. We will look at these real-time control measures in greater depth in Part 2.

■ Ensuring and verifying reliability

For social infrastructure, ensuring reliability is the most critical challenge. Social infrastructure directly contributes to our ability to live stable and secure lives, so many of the systems used in this social infrastructure have no tolerance for outages or interruptions—so-called "down time." We ensure reliability by leveraging the features of both hardware and software. For example, in systems with redundancy, we test that the systems will correctly fail over* if there is a problem. Furthermore, to ensure that devices can deal with any and all irregularities, we not only perform real-time performance testing but also carry out a wide range of tests, including tests of abnormality detection functions and of the reliability of data in the event of a sudden reset.

* Fail over: When part of a system experiences a failure, its functions are automatically taken over by a separate system or component.

Figure 3 shows results from Linux data reliability testing. There is not just one way to save data with Linux. Instead, there are multiple file systems, each with their own characteristics. To select the right file system that best matches the features of the storage, we performed testing that simulated a loss of power and evaluated the data loss that resulted. As the figure shows, results varied significantly depending on the file system's characteristics and configuration.

■ The challenges of long-term maintenance and outside activities related to taking on these challenges

The embedded systems used in social infrastructure must be usable for long periods of time, on the order of 10 or 20 years. Maintenance must therefore be provided for the software embedded in them for as long as the system is used. However, Linux kernels are typically only supported for a short period of time, anywhere from six months to around five years. To ensure stable operation over the long term, security vulnerabilities and bugs must be patched on an ongoing basis, and the safety and reliability of systems must be maintained.

The way the world thinks about long-term maintenance is changing. In October 2024, Europe passed the Cyber Resilience Act (CRA), which aims to strengthen organizations' resistance to cyber-attacks and digital crime and their ability to recover from attacks. Businesses will have to start complying with the first set of CRA obligations starting in September 2026. To respond to these changes, Toshiba, together with Siemens, launched an open source project, the Civil Infrastructure Platform (CIP) Project, in 2016. We are collaborating with various companies to develop and supply software stacks for use in systems that require high reliability and long-term maintenance. One part of those stacks is a Linux kernel that can be supported over the long term, for a decade or longer. We will discuss the activities of the CIP in greater depth in Part 3.

In this part of the running feature, we have looked at the main challenges involved in using embedded Linux in products and the measures Toshiba is using to take on those challenges. Toshiba is continuing to engage in technological innovation to take on challenges like the ones presented here—improved energy-saving performance, shorter startup times, real-time control, long-term maintenance, and ensuring high reliability. The results of our efforts have created an important foundation that will support the future of social infrastructure.

In the next part, we will focus on real-time control and delve deeply into the technical details of how we are tapping the full real-time potential of Linux. Don't miss it.