USB Benchmarks

These benchmarks provide a rough estimate of how USB devices perform on various hardware and software configurations.

I see two possible bottlenecks when bechmarking read/write operations on a storage device: the CPU and the device itself. * If you expected notable differences between the cpus, you would conduct a series of benchmark with the same storage device but different CPUs. You would also make sure that the storage device used in this benchmark is NOT a bottle neck. * If you expected notable differences between the storage devices, you would conduct a series of benchmarks with the same CPU, but different storage devices. Again, you would make sure, that the CPU is not the bottle neck! * there is of course a third possible bottle neck: the USB controller (MAC+PHY) on the SoC and/or possibly the USB-to-SATA converter, in case you use a SATA hard disk. I do not know, about notable differences in CPU, but there definitely are very slow flash storage devices purchasable out there with maximum read/write “speeds” of < 2MB/s and less!!!

• Install benchmark utilities 'hdparm' and 'bonnie++'

  opkg update
  opkg install hdparm bonniexx

• Install USB drivers

   opkg install kmod-usb2 kmod-usb-storage   kmod-fs-ext4  block-mount

• Run the followings commands:

  hdparm -Tt /dev/sda#

  mkdir -p /mnt/benchmark

  mount -O rw,noatime -t ext4 /dev/sda# /mnt/benchmark
  bonnie++ -n 0 -u 0 -s 64 -f -b -d /mnt/benchmark

• hdparm and bonnie++ are non-destructive benchmarks
• bonnie should auto-detect RAM size but you may need to use '-r' parameter to override. Also you may adjust '-s' to be double of '-r'.

Sequential Output = Write performance
Sequential Input = Read performance
Random Seeks = IOPS

• Run the followings commands:

  hdparm -Tt /dev/sda#

  mkdir -p /mnt/benchmark

  mount -O rw,noatime -t ext4 /dev/sda# /mnt/benchmark
  bonnie++ -n 0 -u 0 -s 64 -f -b -d /mnt/benchmark

• hdparm and bonnie++ are non-destructive benchmarks
• bonnie should auto-detect RAM size but you may need to use -r parameter to override. Also you may adjust -s to be double of -r.

Sequential Output = Write performance
Sequential Input = Read performance
Random Seeks = IOPS