Diskgraph
Version: GIT Web: https://github.com/ stolk/diskgraph
Monitoring hard drive performance is an important part of system admin, regardless of the actual hardware format (rotational or solid state). There are plenty of open source tools that are up to the job, but at the same time it’s hard to give an example of a small, lightweight and non-obscure utility that can deliver good monitoring results without hogging system resources.
Diskgraph is one such tool. It compiles in just a few seconds into a tiny 30KB executable, which is then ready to go. Its operation is so simple that you can even omit the full path of the target drive you want to monitor and simply throw in only the last part, just like this:
$ diskgraph nvme0n1
As expected, Diskgraph draws a graph and puts coloured dots on it to represent read and write operations based on /sys/block/nvme0n1/stat data. The horizontal axis is for time, while the vertical axis is for both read/write speeds (left) and number of operations (right). The lowest part of the window displays the legend and the drive model. The latter is perhaps the most important and helpful design decision because it reminds you of the device that you’re currently monitoring.
The green dots are for reading, and the red dots are for writing. Launch Diskgraph against one of your disks and load it with something. For instance, updating the system via its package manager spawns a wave of write operations, whereas searching for a file or running tools such as Disk Usage or Filelight are read-intensive.
Diskgraph isn’t just a toy tool, of course. Use it to monitor remote servers to find out if their disk I/O is the source of slowdowns or other issues. When used over a network, Diskgraph consumes a lot of its bandwidth due to constant graph re-draws, although you can make the terminal window smaller to partially offset this effect.