Technical Success Stories: Understanding Host Usage for Industrial SD Cards
“Our application is not write intensive. We need help finding the best industrial SD card.”
Background
A common question submitted to the Delkin Customer Engineering Team is about selecting the best industrial SD cards for specific host requirements. One of these questions came in from a new customer and they wanted to know what SD card is best for an application that is not write intensive. Despite early discovery efforts, the customer either did not know, or could not share, the usage model or specifics of the application, size and frequency of files written, or any data about the required operating temperature range. The customer was requesting the most cost effective SD card considering the limited number of writes needed during the card lifecycle. While the customer initially requested a TLC based solution, as other bidders were also sampling TLC flash solutions, Delkin recommended a Commercial Utility SD solution. The Commercial Utility SD solution features managed configuration.
Customer Story
The qualification process for the samples submitted from various flash storage manufacturers was nearing an end, and unfortunately every SD card that had been sampled did not pass initial testing. The customer was beginning to get nervous as they thought they had plenty of time to choose the right SD card prior to the product launch date, but as of that time there was not a single SD card approved for use in their host.
Because multiple brands and samples of cards had failed, the customer concluded that the problem did not lie in the cards alone. The focus shifted to analysis of the host’s usage model during qualification testing. During that process, the customer tried to replicate a real-world usage model by cycling temperature up to 85C and writing a sample of data the host was designed to record. They did this continuously, not knowing that this could cause a problem.
Delkin Discovery
Delkin’s Customer Engineering Team researched this further, first analyzing the SMART data on the failed SD cards that had been returned after qualification testing. It was clear to see that the program/erase cycle counts performed during testing were well beyond the flash rating. All spare blocks had been spent, rendering the card in a read-only state. After supplementary examination, Delkin’s Customer Engineering team discovered that the application workload model consisted of negligible writes (less than a page) to multiple partitions. This is considered random small writes, which increases write amplification. Because of the customer’s write-intensive qualification process, they had written more than 9TB of data to the MLC flash, wearing out the card in a relatively short amount of time. The kind of cards sampled were also prone to high temperatures, which made the failure occur faster. While the Commercial UtilitySD card outlasted the competitors’ TLC cards, it still did not meet field requirements.
This situation happens when host OEMs are not fully aware of the size of files being written, frequency of writing, and the true operating temperature, yet these are all very important factors to consider. Delkin has the capability of using advanced testing equipment software programs to help determine, replicate, and analyze the real-world usage of each customer’s host application, and point towards the most ideal solution.
Delkin Solution
After further analysis of the SD cards, Delkin’s Customer Engineering team proposed testing a lower capacity SLC card, which would have the advantage of a much higher program/erase cycle rating. In addition, due to smaller page and block sizes, there is lower write amplification than a MLC or TLC solution. The customer performed another round of qualification testing with the new Industrial SLC SD card samples provided by Delkin and was able to get the samples to pass in time to meet the target launch date.
While the SLC SD card solution was more expensive than the TLC SD solution that the customer had originally requested, it performed flawlessly under the actual usage scenario, and enabled the customer to hit their new product launch date and begin collecting immediate revenue. By switching to a lower capacity SLC based card, the TBW was drastically improved. Delkin was also able to train the customer on “preventative maintenance” using the card’s SMART data. This data then helped them to make updates to future versions of the host’s software to improve efficiency on how the host wrote data to the flash. This customized lifecycle management worked to improve the longevity of the card lifecycle.
Lesson Learned
It’s important to remember that a NAND flash memory card is a very important component in ensuring the host application functions as intended. Industrial SD cards are more complex and customizable and should be considered more than just a “memory chip”. In this scenario where small writes resulted in high write amplification to the flash, spare blocks were retired earlier than necessary, and the high temp testing accelerated these failures. By simply switching from a MLC or TLC card to an SLC card, program/erase cycles were increased by at least 20 times. This exercise also helped the customer discover ways the host program could be modified to become more efficient with the flash.
Lean on Delkin’s Technical Team
Have you had a problem with your host storage? The Delkin Customer Applications Team stands at the ready to be your trusted advisor for your host critical storage. We have solved hundreds of complicated host failures and look forward to understanding your usage model for flash based Rugged Controlled Storage.