Survey Results on Real-time OS Use in Embedded Systems

Once each year the TRON Association conducts a questionnaire-based survey on trends in real-time OS use in embedded systems, for reference in planning the future course of the ITRON subproject. Reported here are the results of the latest survey, carried out from late 1998 to early 1999. This time a survey on GUI use in embedded systems was conducted along with the real-time OS survey. The results of the GUI survey are reported separately and are therefore not indicated here.

1. Outline of the Survey

Areas covered by the survey included the status of embedded system development, problems with a real-time OS and criteria used in selecting one, use of the ITRON-specification OS and user impressions of it, awareness of ITRON-related activities, and wishes for the project. Similar surveys have been conducted since 1996, making this the third time this particular survey has been carried out. As noted above, this time a survey on GUI use in embedded systems was conducted at the same time.

• Results of the last year's survey (survey period: Nov. 1997 - Jan. 1998)
• Results of the previous year's survey (survey period: Nov. 1996 - Jan. 1997)

1.1. Survey period

The survey was conducted from November 1998 through January 1999.

1.2. Survey method and response rate

(1) How the survey was carried out

Questionnaires were distributed by the following means, then were collected and the results were tabulated.

• Distributed at an embedded systems trade show
• Sent by direct mail

The trade show was the Microcomputer System & Tool Fair '98 (MST '98) held in November 1998 in Tokyo, where questionnaires were distributed at the TRON Association Booth and at the seminar hall. Where possible, the questionnaires were filled out on the spot and collected immediately. Direct mailings were sent mainly to those who took part in the previous survey. All the direct-mailed questionnaires were sent to addresses in Japan.

(1) Number distributed and responses

Of the questionnaires distributed by the above means, 678 responses were received. This is approximately 50 percent higher than the previous survey (438 responses) and 2.3 times greater than the 287 responses received for the first survey. This means the rate of increase has been around 50 percent each year. The number of questionnaires of each kind and the response rates are as follows.

The response rate for the questionnaires distributed at the trade show was much better than the previous year's 19.9 percent, largely as a result of asking that they be filled out immediately. The response rate for questionnaires distributed by direct mail was nearly the same as in previous years (25.0 percent in the first survey and 24.2 percent in the second). Accordingly, for the first time a higher return rate was received for questionnaires passed out at the trade show than for direct mail.

1.3. Surveyed population and survey items

Given the nature of the survey, the questions were directed primarily at developers of embedded systems, who are either users or potential users of a real-time OS. The main items surveyed are as follows.

(1) Recently developed embedded systems

The survey asked about the application fields, system scale (processor type and program size), programming languages used, and real-time OS used in up to three embedded systems recently developed by the respondent.

(2) Difficulties with real-time OSs and basis for selection

These survey items asked the respondents to rank the importance of problems with using a real-time OS in embedded systems, and of real-time OS selection criteria.

(3) Awareness of ITRON-specification OS and related activities, etc.

The respondents were asked about their involvement with the TRON-specification OS, and the extent of their awareness of ITRON-related specifications, committee and study group activities, promotional efforts and the like.

(4) Advantages and disadvantages of an ITRON-specification OS

Persons using, developing or considering the use of an ITRON-specification OS were asked about its advantages and disadvantages.

(5) Wishes for the ITRON subproject

Finally, the questionnaire offered respondents a chance to express their desires regarding ITRON-related activities.

1.4. Occupations of the respondents

The number of respondents saying they were in design or R&D amounted to approximately 90 percent of the total, which means the survey reached its target audience.

2. Analysis of the Survey Results

2.1. Method of tabulating the survey results

In tabulating the results of the survey, where answers were given that could not be interpreted as correct responses, as a rule only the affected item was treated as invalid. For example, when multiple answers were given where only one was to be selected, that item was ruled invalid. In the case of questions allowing for multiple answers but requiring that the choices be ranked by entering "1" for the most applicable, then "2" and "3" for the 2nd and 3rd in applicability, the response was ruled invalid if this method of ranking was not followed, e.g., when "1" was entered for more than one choice.

In tabulating the results of ranked multiple-answer responses, the percentage of "1" rankings for a choice is indicated (under the category, "single responses"), and also the percentage of "1" or "2" or "3" rankings for a choice ("multiple responses"; thus no distinction is made between "2" and "3" rankings). Since the total of "single responses" is 100 percent, these results are shown in a pie chart. Totaling the "multiple responses" results in a percentage greater than 100 percent, so these results are shown on a bar graph. In so doing, a distinction is made between choices ranked as "1" (single responses) and choices ranked as "2" or "3" (multiple responses).

The population in the following results is the number of respondents in all cases except 2.2, where it is the number of embedded systems.

2.2. Recently developed embedded systems

The respondents were asked about five items - the application fields, processor type, program size, programming languages and real-time OS - used in a maximum of three embedded systems recently developed by them or their company.

Valid responses were received for an average of 1.5 embedded systems per respondent. Note, however, that if an invalid response was received for any of the five items asked about, the entire response (for all five items) was considered invalid. This approach was taken for the sake of cross-tabulation of results, even though another option would have been to exclude only the invalid parts in the item-by-item tabulations.

The valid results indicated here are for 945 systems. The percentages given in the results for this section are all based on this population (number of embedded systems), which represents an increase of approximately 1.4 times the number of systems covered in the previous survey (675 systems).

(1) Application fields of embedded systems

Consumer-oriented products (home appliances, audio/visual, entertainment and education, personal information appliances, and communication equipment (terminals)) accounted for around 35 percent of the total, while manufacturing equipment (plant control, factory automation) totaled approximately 20 percent, with other kinds of industrial systems making up the remainder. These figures can be regarded as a reasonably accurate breakdown of embedded systems developed in Japan, considering that they represent the number of systems developed, not the number produced. This is because consumer products tend to be produced in large volumes but the number of different systems developed is not as large.

(2) Processors used in embedded systems

In the recently developed embedded systems surveyed here, 32-bit processors occupied the largest share at around 47 percent, followed in order by 16-bit and 8-bit models, indicating that 32-bit processors have become the mainstream choice even in the embedded control field.

Compared to the previous survey, the use of DSPs and other specialized processors has nearly doubled, although their actual number is still quite small. Smaller increases were registered by 8-bit, 16-bit and 64-bit processors, with a slight decline in 32-bit systems.

In analyzing these results it should be noted that respondents were asked to choose up to three of the most recently developed systems, in a survey on real-time OS use. If respondents were involved recently in more than three system development projects, the natural tendency would be to choose systems of a scale that would likely use a real-time OS. Also, if one system uses several processors, the responses concern only the main processor, so that in reality 4-bit and 8-bit processors are probably in wider use than these results suggest.

(3) Program size in embedded systems

As in previous surveys, program sizes were divided rather evenly into the four categories of 64 KB or less, 64 KB to 256 KB, 256 KB to 1 MB, and 1 MB or more. For the first time the survey further sub-classified the systems above 1 MB, with the result that large-scale programs of 16 MB or above were reported for 3.6 percent of the systems. Programs of this size were used mainly in the plant control/factory automation field and telecommunications field (network equipment).

Compared to the previous year there was a slight decrease in the 64 KB to 256 KB category, and an increase in programs above 1 MB. The increase in programs above 1 MB is even greater in comparison to two years ago, showing the trend toward larger scale in embedded software.

(4) Programming languages used in embedded systems

The programming languages in main use (single responses) are C language at nearly 70 percent and assembly language at just over 20 percent of the total. When multiple responses are looked at assembly climbs to more than 45 percent, indicating that in many cases C is used as the main language and assembly for certain parts.

C++ was named in approximately 7 percent of the multiple responses, about the same as last time. Java tallied less than 1 percent of the total, suggesting that the use of Java for programming embedded systems is still rare. Other languages receiving multiple mention were Visual Basic, PL/I and FORTH.

(5) OS embedded in embedded systems

The overall trend in OS use this time shows that 28 percent of the systems reported on use no OS, approximately 31 percent use an ITRON-specification OS (broken down to 19 percent commercial ITRON-specification OS and 12 percent in-house ITRON-specification OS), 17 percent use an original in-house OS, and 26 percent of the systems use a non-ITRON commercial OS. Of the systems that use an OS, around 43 percent use an ITRON-specification OS; and of the commercial OSs used, around 44 percent are ITRON-specification OSs. Of the systems not using any OS, in around one in eight cases (3.5 percent of the total) the reason for this choice was said to be "Due to problems" preventing its use.

Standing out among the non-ITRON commercial OSs were the Microsoft OSs (MS-DOS, Windows 3.1/95/98/NT/CE) or compatibles at around 9 percent, and Wind River Systems products (VxWorks, etc.) at around 4 percent. All other OSs were at less than 2 percent each, indicating how fragmented the real-time OS market has become. The OSs coming in at less than 1 percent of the total were all lumped together in the results under "Other commercial OS."

The share of the ITRON-specification OS is up slightly from the previous survey, as well as from the one before that, suggesting that the use of the ITRON-specification OS is still on a rising trend. Among ITRON-specification OSs, those available commercially enjoyed increasing use, while there was a decline in in-house original-implementation ITRON-specification OS use.

Due to an error in preparing the questionnaire, in-house ITRON-specification OSs and BTRON-specification OSs were inadvertently placed together as a single choice. Since the number of cases of BTRON-specification OS use in embedded systems is very small, this error is unlikely to have influenced the overall trends. Moreover, the latest BTRON-specification OS adopts the ITRON-specification OS as its core and can in that sense be classified as an ITRON-specification kernel.

The wording on the questionnaire was changed from "OS used" on the previous survey to "OS embedded" this time. The reason for this change was to avoid confusion with the OS used on the host machine for development. The decline in Microsoft OS figures from the previous survey can possibly be traced to this influence.

The relation of application field to embedded OS use was analyzed. Graph 8 arranges all the embedded systems appearing in the survey according to their application field, and indicates the OS use rate for each. Graph 9 shows the same results when systems not using an OS are excluded (the number of valid responses for this graph is 682). In these graphs, except for "Other" the application fields are arranged from right to left in ascending order of ITRON-specification OS use in that field. The numbers at the top of the bars are the sample size for each application field.

While some application fields have only a small sample size, as a general trend ITRON-specification OSs have a greater than 30 percent share in consumer-oriented systems (personal information appliances, communication equipment (terminals), audio/visual equipment, entertainment/education, and home appliances). When systems not using any OS are excluded, the ITRON-specification OS share in these fields is above 50 percent, strong evidence that the ITRON-specification OS is being widely used in consumer products. The share in the communication equipment (terminals) field is also more than 40 percent. In this field, however, systems not using an OS are few. As a result, the results for the ITRON-specification OS in Graph 9 (only systems using an OS) for this field are not much different from the Graph 8 results.

Compared to the previous survey, the share of the ITRON-specification OS in fields where it traditionally does well shows an upward trend, whereas in other fields (in particular those of commercial equipment, personal computer peripheral/office equipment, and plant control/factory automation) the share has dropped. An exception is the field of transportation-related systems, where the ITRON-specification OS has increased its share nearly four-fold since the previous survey. It is uncertain whether this result is related to recent efforts in the ITRON subproject to study needs in the automotive control field, but there is little doubt that awareness of the ITRON-specification OS in this field has risen. The transportation equipment field is one in which the ratio of systems not using an OS has decreased, as OS use has become more common in this field. The use in this field of the OSEK/VDX-specification OS, which is specialized to this field, was mentioned only three times; but inasmuch as that OS is attracting international attention, it will be interesting to watch this trend in the future.

Looking at the relation of processor scale to OS used, the following trends can be seen. First of all, there were no examples of OS use with a 4-bit processor. Around 30 percent of 8-bit systems use an OS, with most of these being an in-house original. Of 16-bit systems, more than 30 percent use no OS and another 20 percent use in-house original OSs, these two categories making up half the total. Of 32-bit systems, around 10 percent use no OS and slightly more than 10 percent use an in-house original OS, together making up less than a quarter of the total. Both the share of the ITRON-specification OS and that of other commercial OSs here is around 40 percent each.

Compared to the previous survey, the use of an OS in 8-bit and 16-bit systems has declined, and so has that of the ITRON-specification OS in these systems. OS use in 32-bit systems is about the same as the previous year, with the ITRON-specification OS share showing an upward trend here. This would indicate that the ITRON-specification OS is being applied increasingly to large-scale systems.

Finally, program size relates to embedded OS use as follows. Of systems with a program size of 64 KB or less, more than 60 percent do not use an OS, around 20 percent use an in-house original OS, and the TRON-specification OS is used in approximately 10 percent. This trend closely matches that of 8-bit CPUs. The percentage of systems with no OS goes down as program size increases. The ratio of ITRON-specification OS use increases with larger program size, peaking at around 45 percent in the 1 MB to 4 MB range. In programs larger than that, ITRON-specification OS use decreases. These results indicate that the ITRON-specification OS is most widely used in systems of medium scale and lower.

Compared to the previous survey results, the use of an OS in smaller programs, under 64 KB, has dropped, as has use of the ITRON-specification OS. In systems larger than that, the OS use rate is about the same and the ITRON-specification OS share has increased. From the standpoint of program size, then, application of the ITRON-specification OS is gradually shifting toward larger systems.

2.3. Real-time OS difficulties and selection criteria

All the survey subjects were presented lists of possible difficulties with using a real-time OS and of OS selection criteria, and were asked to choose up to three applicable items.

(1) Difficulties with using a real-time OS

The answer chosen most often as a problem with real-time OS use was "An absence or shortage of staff familiar with the technology." Approximately 54 percent of the respondents chose this as a problem and around 33 percent selected it as the most important issue.

In the single-response totals, the next most common difficulties selected were "Cost is too high" and "Lack of a development environment and tools" at around 15 percent each. These were followed by "Major differences in OS specifications, making it hard to switch," "Performance and functions do not meet our requirements," and "OS size or resources used are too large," each chosen by between 7 and 9 percent of the respondents. Looking at multiple responses, "Lack of a development environment and tools" at around 42 percent and "Cost is too high" at 36 percent stood out from the others. The issue of development environment and tools is not the biggest issue but can be considered next in importance.

The overall trends are unchanged from the previous survey. There is a slight increase in "Cost is too high" among the single responses and a slight decline in "Lack of a development environment and tools," reversing their order from last time. "An absence or shortage of staff familiar with the technology" showed an increase, while most of the remaining choices declined somewhat.

When those who chose "Lack of a development environment and tools" were asked to name the specific tools lacking, many cited debugging tools (debuggers, ICE, etc.). The lack of a task-aware debugger was frequently noted. Other commonly listed tools were integrated development systems, CASE tools, and tools for system verification.

(2) Real-time OS selection criteria

In the single-answer results, the selection criterion most often cited was "Performance and functions match our requirements"; however, this trails behind the combined total for two related items, "Has a proven track record in our company" and "Is widely used in the industry," both focusing on the installed base and together making up around 33 percent of the responses. "High reliability" and "Low cost" were next at just over 10 percent. "Good supply of software components" was selected by a mere 1.5 percent of the respondents, indicating that this factor is not given much weight in choosing a real-time OS.

When multiple answers are considered, "Low cost" appears more often than among the single answers. The results would suggest that while cost may not be the most important factor in selecting a real-time OS, it is considered second or third in importance.

The rank among single responses differs little from the previous survey, but there was a significant drop in the percentage for "Performance and functions match our requirements" and a rise in some other items, notably "Is widely used in the industry," "High reliability" and "Low cost." Of the items relating to installed base, "Is widely used in the industry" rose while "Has a proven track record in our company" dropped somewhat, indicating the growing importance placed on de facto standards. "Low cost" as a selection criterion showed a notable increase especially among multiple responses, perhaps influenced by the protracted economic slump in Japan.

2.4. Awareness of the ITRON-specification OS and related activities

(1) Involvement with the ITRON-specification OS

The respondents who claimed to "Have used or developed an ITRON-specification OS" and to "Have investigated or considered its use" together made up more than 70 percent of the responses, whereas "Was not aware of ITRON" was selected by a mere 0.3 percent, underscoring the high degree of ITRON specifications awareness.

The response rates for both "Have used or developed an ITRON-specification OS" and "Have investigated or considered its use" were up slightly over the previous survey. Continued promotional efforts will need to be directed at those who responded "Have heard about ITRON," encouraging them to investigate and consider adopting the ITRON specifications.

(2) Awareness of ITRON specifications

Awareness of the mITRON3.0 specification rose nearly 20 percent since the last survey, to approximately two thirds of the respondents. This level is closely matched to the number of respondents claiming either to "Have used or developed an ITRON-specification OS" or to "Have investigated or considered its use"; it also shows an advancement from merely knowing of "the ITRON specification" to being aware of the version number as well. Even the mITRON4.0 specification, which was a new item on this survey but had not yet been released, was known to one in four respondents. This would seem to indicate that efforts to promote the ITRON subproject are having a certain amount of success.

(3) Awareness of ITRON-related committees and study groups

The respondents to this survey tended to be more aware of ITRON-related committees and study groups than on the last survey. Awareness of the ITRON Technical Committee was up to nearly two thirds of the respondents.

(4) Awareness of ITRON-related PR

The percentage of respondents who were aware of the ITRON Web site increased over the previous survey, but there was a drop in awareness of "Booths or presentations at trade shows and exhibitions" and of the ITRON Newsletter.

(5) Awareness of other ITRON-related activities

Although awareness of the Registration System for ITRON-specification Products was up from the last time, overall the awareness of activities listed under this category was not very high.

2.5. Advantages and disadvantages of the ITRON-specification OS

The respondents who answered either "Have used or developed an ITRON-specification OS" or "Have investigated or considered its use" under 2.4 (1) above were next asked about the advantages and disadvantages of this OS. Presented with a list of items, they were asked to choose up to three that applied.

(1) ITRON-specification OS advantages

The advantage, "The specifications are easy to understand" was cited at a very high rate, by 30 percent of the respondents in the single response results and around 47 percent in the multiple responses. This result points to the success of the ITRON specifications design policy with its emphasis on engineer training ease. The next most common responses were "Small OS size and resource use," "Supports a wide range of chips," and "Low cost." Similar trends are to be seen in both the single and multiple response results. An advantage often cited under "Other" was the open specification policy.

Compared to the previous survey, the response rate of "The specifications are easy to understand" increased, "Supports a wide range of chips" and "Low cost" were up slightly, while "Small OS size and resource use" decreased. One reason for this trend may be that size and resource use are becoming less of an issue with the rising curve of processor performance and memory availability.

(2) ITRON-specification OS disadvantages

"Lack of a development environment and tools" was still the most commonly cited disadvantage, accounting for around 23 percent of the single response results and registering around 36 percent in the multiple responses. This percentage has gradually decreased, however, over the three surveys as progress is made in filling the needs for ITRON-specification OS development systems and tools.

Next in frequency was "No notable disadvantages," at 17 percent of the single responses, followed in order by "Too many implementation-dependent parts making it hard to port," "Lack of software components," and "Lack of engineers familiar with it." In the multiple responses, "Lack of software components" received a relatively high rate of response. Among the issues cited under "Other" was the low awareness of ITRON overseas.

In comparison to the previous survey results, decreases were to be seen in "Lack of a development environment and tools" and "Lack of engineers familiar with it," whereas the response rate increased for "Too many implementation-dependent parts making it hard to port" and "Lack of software components." These results would seem to confirm the correctness of current efforts in the ITRON subproject toward greater standardization, aimed at improving the development tool lack and the two disadvantages that showed an increase on this survey.

2.6. Wishes for ITRON-related activities

All the respondents were asked to select up to three areas in which they would most like the ITRON subproject to become involved.

The rate for "Interface standards for software components" increased over the previous survey, taking over the top spot from "Standardization of interfaces with development environments," which decreased. This result is congruent with the decrease in citing of "Lack of a development environment and tools" as an ITRON-specification OS disadvantage, even though that is still seen as a significant issue. The rate for "C++ and Java language binding standards" was down, dropping that item two slots in the ranking. This may be a result of adding "Support for Java runtime environment" as a choice. The other items showed a downward trend.

Chart - single
Chart - multiple

Sheet 20 : Wishes for Further ITRON Involvement (single and multiple responses)

Wishes	Response rate
	Single	Multiple
Interface standards for software components	25.3%	46.6%
Standardization of interfaces with development environments	22.1%	44.8%
Free ITRON-specification OS	10.9%	25.8%
Network support	9.5%	25.8%
C++ and Java language binding standards	8.6%	20.2%
Application design guidelines	4.1%	15.9%
hard real-time support	4.0%	13.5%
Educational seminars	3.4%	11.2%
Java support	3.1%	10.3%
Standardization geared to certain application	3.1%	8.1%
Multiprocessor support	2.5%	6.2%
Fault tolerance support	0.5%	4.4%
Other	1.4%	2.5%
None of the above	1.5%	1.5%

Valid answers : 651

3. Summary

The questionnaire was similar to that used for the previous two surveys, with some changes and additions. The 50 percent increase in the number of responses raises the statistical relevance of these results. While there were no major changes from the overall results last time, some interesting trends can be discerned by looking at the results over the course of the three years this survey has been taken. In particular, use of the ITRON-specification OS continues to rise, being selected for more than 40 percent of the systems that used an OS. This is far more than any other OS, providing further evidence that this OS specification has become an industry standard in Japan.

Generally the results are favorable for the ITRON subproject, but the number of persons citing the lack of development systems and tools continues to be high. Another notable trend is the increase in response rate for poor portability and lack of software components. That these are issues currently being addressed in the ITRON subproject indicates that this is the correct course to be taking, but the results also show the importance of devoting even more of an effort to overcoming these weaknesses.

The survey results this time for real-time OS problems and selection criteria show that development environment and cost have gained in importance, while more technical factors have declined relatively. For one thing this can be attributed to the maturation of real-time OS technology, while the economic situation in Japan may be another factor.

The TRON Association plans to reflect these results in shaping the future direction of the ITRON subproject, while continuing to keep track of trends through similar surveys in the years to come.

References

• Results of the last year's survey (survey period: Nov. 1997 - Jan. 1998)
• Results of the previous year's survey (survey period: Nov. 1996 - Jan. 1997)
• The Collection of Charts (PDF Format)

---

Back to ITRON Survey Activities (in Engish)