Category Archives: 03. Sw design

SW partitioning – why do you try to implement it?

SW partitioning is seems to be a very useful approach to decompose ASIL in SW. I understand that if it is implemented, some SW parts can be excluded in the FuSa scope. But I have a different point of view. My question is that is it really give us benefit to reduce our resources?

Regarding to applying SW partitioning, its motivation is to reduce resource. It was my involved project’s interest. If some SW components are excluded as QM, then we have reduced scope for FuSa, and it is benefit to us. When I heard its trials, it sounds good, but I realized that its result would not be good. I’ll tell you what should be considered additionally.

Let’s assume that we should apply a watchdog. It requires many technical consideration and it does not mean just add watchdog in a circuit. Even if watchdog is added, main purposes cannot be achieved if it is not implemented smartly. (reference; proper watchdog timer use in Korean)

Similarly if we implement SW partitioning, design should be verified. At least, task scheduling plan should be specified in a SW design, its design should be reviewed, and its effectiveness should be tested during or after SW integration.

if critical faults occur in a QM side, it shall be ensured that OS catches such exceptions and manages as we intended. it is one side of SW partitioning.

The other consideration is a space partition. A SW component in one partition can try to access memory region or resources that are allocated to other partition. In this case, it shall be ensured that such trials are not succeed. And such trials are invalid, exceptions or any error events should be triggered. Then exception handling should be considered. How will you do if such exceptions are occured? SW reset? Are you sure?

For now, I don’t think that SW partitioning will save our resource. It is a safety design concept, not a method to save our resources. But I may change my mind if I will experience more, but this idea is my latest idea for SW partitioning.

Software partitioning in AUTOSAR

software partitioning is a development concept to decompose ASIL. It is like a firewall that protects a failure propagation from the other partition. In some ways, it is very close to ASIL decomposition concept used in the SW part, so its concept is recommended to consider it also.

To implement the concept, supporting mechanisms shall be supported. As I know, existing version of AUTOSAR cannot support it. Maybe it will take some time to use this scheme. The reasons that I thought are that there are many approaches suggested in the paper to improve partitioning in AUTOSAR. Such studies reflect that this is in progress, and needs more time to complete.

And when I attend Vector conference, I heard that there will be a need for very powerful computing for a ECU.

When the day has come, it will cover many features. In this environment, SW partitioning will be very useful technology. For now, it seems to be immature for AUTOSAR to support SW partitioning scheme.

If what I know is incorrect, then please give a feedback as a reply.

Does usage of modeling tool achieve requirement for Semi-formal notation?

Of course not. I’ve experienced that they use Simulink tool as a drawing tool. They really spend too much for drawing tool. They should choose visio. It is great for drawing. Their Simulink modeling was not met for syntax and semantics.

I’d like to ask. Do you really think that usage of such an expensive tool can achieve semi-formal notation requirement?

Specifying semi-formal notation means that the design explains itself and it is able to  implement. But what if there are syntax errors? Then, it does not meet for semi-formal notations requirement. The design cannot explain itself. There maybe several ways of interpretations. But it is not the intention of semi formal notation requirement.

If you try to use UML diagram, then please consider syntax and semantics. It is really important. Do not use it just for “drawing tool”. You are not an artist. and it cannot be an art. Please do engineering, not art.