Percepio Device Firmware Monitor™ – The Next Step in IoT Software Quality Assurance
Let’s face it – we can never be certain that any software is free of bugs. On average, 5% of all the bugs introduced during embedded software development actually remain in the production firmware, despite all verification efforts. For most products, it is practically impossible to test every possible usage scenario and code path – they are simply far too many. You can always spend more time and money on verification, but the effort needed to find the last remaining bugs tends to increase exponentially and development projects have time and budget constraints to meet.
Another concern is validation. Even if you are testing all code according to best practices and everything seems to work perfectly, the requirements and test cases might not reflect how your customers will be using the product in practice.
The real test comes when thousands of people start using your product, in ways you never anticipated. Will it stand the test of reality?
Brief demo showing how an automatic DFM error report is received via email and then opened in Tracealyzer (full-screen recommended!)
Missed bugs may at the very least irritate your customers, damage your reputation and hurt sales. In some cases, bugs may even lead to accidents, product recalls and legal action. The rise of Internet-of-Things (IoT) makes firmware quality assurance even more challenging, but IoT also provides a new remedy – Over-the-Air (OTA) software updates.
If you fix the problem and update the shipped devices quickly, your customers will be more satisfied and might not even notice the issue. However, you cannot fix bugs that you are not aware of. Automatic feedback is needed to make developers aware of firmware issues in deployed products.
Automatic Feedback Within Seconds
Enter Percepio Device Firmware Monitor (DFM), a ground-breaking new cloud service for IoT product organizations that provides awareness of firmware problems in deployed devices and speeds up resolution. When a firmware issue has been detected, DFM notifies the developers within seconds and provides diagnostic information about the issue, including a trace for Percepio Tracealyzer. This shows what was going on in the code when the error occurred, making it far easier to understand the problem and quickly find a solution.
“Percepio’s Device Firmware Monitor is a game-changer in that it enables instant feedback from systems deployed in the field, to ensure your firmware quality is constantly improving.”
Jack Ganssle, Principal Consultant, TGG
“Percepio’s Device Firmware Monitor is early to the market and original. IoT developers need this sort of direct feedback from their deployed systems.”
William E. Lamie, President, Express Logic
Without automatic feedback, you actually rely on your end users to report any issues, a responsibility they have not agreed to. Then you might not hear about the issues until it’s too late, when many customers have already been affected. Moreover, your end users can’t be expected to provide sufficiently detailed information for you to quickly identify and solve the problem. A vague error report like “the screen went black” may require weeks of guesswork until you find a likely cause, and even then, you still don’t know if really you solved the right problem. Imagine how much troubleshooting time that could be saved if you instead had access to detailed diagnostic information about every issue in the production software.
DFM is designed to leverage existing secure solutions for cloud connectivity, storage and OTA updates. Percepio DFM initially supports Amazon Web Services (AWS IoT Core) and Amazon FreeRTOS, but support for additional platforms is planned and can be provided on request.
The information flow starts in the error handling code of the IoT device, such as sanity checks and fault exception handlers. By calling the DFM firmware agent from these locations, firmware issues are uploaded as “alerts” to the customer’s cloud account. Alerts may include an error message and any other information of interest to the developers for the specific issue, such as software state variables and hardware registers. Depending on the severity of the issue, the alert is either uploaded directly or after a device restart, once the cloud connection has been restored.
The alert also includes a trace of the most recent software events prior to the issue, which is recorded automatically by the DFM agent. This tracing technology builds on 15 years of experience in RTOS tracing and is 4-8x more memory efficient than traditional RTOS tracers – only 4 KB is needed to store a trace containing up to 1000 software events. The efficient trace encoding is very important for two reasons – it allows for collecting traces of sufficient length even from memory-constrained IoT systems and it minimizes the cloud-side operational costs of DFM messaging and storage.
The alerts from the DFM firmware agent are uploaded to the customer’s cloud service (AWS IoT Core), which is configured to store the alerts (Amazon S3) and also to notify the Percepio DFM Classification Engine. This service is the core of the DFM solution and runs in Percepio’s AWS account. It is responsible for classification, statistics and notifications to the developers. It also offers configuration options for DFM, e.g., in what conditions notifications should be sent and where to send them.
The software trace never leaves the customer’s cloud account. Only an anonymized signature of the issue is provided to the Percepio DFM Classification Engine and this information is completely transparent and configurable for the customer.
When the developers receive a notification about a new DFM alert, they can access the alerts and traces directly from Percepio Tracealyzer. The DFM Dashboard in Tracealyzer shows the recent alerts and allows for high-level analysis, e.g. if a certain issue was fixed by your latest firmware version. Moreover, the traces can be opened directly from the DFM Dashboard, thanks to Tracealyzer’s built-in integration with Amazon S3.
Cloud-Side Operational Costs
DFM does not generate any data traffic unless an issue is detected, so if you don’t have any missed bugs in your code, there is no DFM activity that drives cost. You need to store the alerts for some time, but due to the small amount of data per alert (typically 5 KB) and today’s cheap cloud storage, this is a negligible cost. Especially when compared to the value of the provided information. Say that you have a large fleet of one million devices, with a lot of firmware issues – say 1 alert per device and week on average, and each alert is 5 KB. This would produce about 260 GB per year. Storing this data for a year would cost about $72, assuming Amazon S3 standard storage. But since most data can be deleted after a short time, e.g. duplicates of the same issue from different devices, the storage needed can be reduced to a fraction of this level.
Sending out OTA updates in response to DFM alerts can be a significant cost for large fleets, but this must be compared to the alternative cost of letting the bug remain unfixed, e.g. damaged customer experience, reduced product sales, or even accidents and legal action. How much would that cost? And, in case of minor issues, you can wait and see if more issues are reported, and then issue a single update with all corrections.
Does this sound interesting? Please contact us at email@example.com. We are looking for early adopters to provide feedback. By joining our pilot program, you get to use the Percepio DFM service for free!