Hello, we are seeing intermittently poor disk (both ram and sd card )performance on our colibri T30 WEC7 BSP 2.3 devices.
Our scenario is that we are receiving many small payloads, around 220 Bytes from a Lora transceiver and trying to send those out as files via Ftp to one or possibly two servers.
Our devices are mostly connected via modem to the Ftp servers in the field so we try and anticipate communications disruptions, to this end, we developed software that will
- take the incoming Lora payloads as files and make a series of copies of the files to time stamped directories
- send those files via Ftp to our servers
- delete the files in the stamped directories to indicate that they have been sent
Consequently, we have quite a number of reads and writes of small files to the disk going on and sometimes we see hiccups or long times (sometimes on the order of a few minutes to make 100 copies and 50 deletes of 220 files) to move OR read these files.
We can see that this is happening if we if read and write the files from RamDisk OR
SDHC (16 Gig, 1 Partition, exFAT formatted, sector size 512 Bytes ) Card.
One possibility is the high CPU usage of core 1 (out of 4 cores) that is starving the threads necessary to perform the disk IO. Or is it the case that the system will use all 4 cores and be able to avoid this problem? We use quite a few apps that use the compact framework and sometimes they are cpu heavy on core 1
We are under a tight deadline to get these times under a minute in the worst case. Do you have any advice on tracking “hiccups” in IO performance?