I'm doing a simple application to log data from DMA channel to USB disk. The data throughput is about 5.5MB/s. I'm experimenting with a 64GB USB disk. I'm noticing a strange behavior. There is a chance for the whole application to go on until the USB disk becomes full. But, for most cases, the application stops randomly during the writing process. By adding some timer in the source code, I find that the USB write operation could suddenly take 1~5 seconds, where 40ms ~ 80ms is expected. CPU usage is not that high, about 60% and 80% for the two CPUs. Some expert tells me that there could be conflicting in Zynq IO. However, most IO monitor commands don't work in NI Linux RT. By using "top", I notice that, right during the failure, the IO usage goes up to about 25~30% and sirq goes up to 25%. Do you have any suggestion on how I can further debug/analyze the issue?