Jump to
DiskSpeed is a Lucee application running in a Docker container that can perform benchmarking of system controllers & hard drives.
(click thumbnails for larger images)
DiskSpeed was designed to run on UNRAID and has support for the UNRAID OS but should also work fine under any unix OS. Windows is not supported as the app requires privileged mode in order to scan the hardware which Windows does not support.
Drive controllers are identified and the ports scanned to see what drives, if any, are attached. The controller can be tested to see if the drives attached are capable of saturating its bandwidth by performing a 15 second benchmark on each drive in sequence, then reading all drives simultaneously and comparing the results. If the simultaneous scan is showing lower read rates, it is possible for it to bottleneck. On the UNRAID OS, performing a parity scan on the UNRAID array involves reading all drives at the same time and thus a controller bottleneck will increase the duration of the scan.
Benchmarking hard drives with spinning platters is done by reading the drive every 10% (default) of the drive including the start & end for 15 seconds giving you a basic speed curve over the surface of the drive. These benchmarks can be compared over time and will indicate if a drive is failing if a benchmark is showing slower speeds than previously. If benchmarking multiple drives at once, only one drive per controller is tested and Solid State Drives will wait until they can be tested alone.
DiskSpeed will only perform read tests on a drive, not write tests.
Benchmark with multiple like models and an unlike model
Parity scans will operate at the speed of the slowest drive in the array. Even though drive "data6 (sdg)" has a clear speed advantage over the other drives, it will not impact any parity scan as it'll be slowed down to match the others.
Viewing an hard drive will display specific information derived from the drive itself. If the drive doesn't reveal it's RPM, a micro-benchmark is done to reveal it. WD-Green drives are known for this. You can compare benchmarks taken at different dates to see if the drive is starting to perform poorly. Such signs of aging might not be evident in the SMART reports but can show up as a dip in a spot on the drive's read rate.
Benchmarking solid state drives can't be done the same way as a hard drive. You can ask to read the drive at a given location but most SSD's will not even attempt to do so if that location has never been written to, leading to impossibly high speed results. In order to test, large files with random data need to be written to the drive and then read back. This could arguably decrease the lifespan of the SSD so be aware of that, though I would also argue that the wear & tear is negligible.
During the write phase, up to 15 sets files are created and the write speeds are compared to the previous ones. Each set of files is comprised of four files that are written simultaneously that combined equal the total requested test set size. For NVMe drives, the set defaults to 4GB and for SATA SSD drives, 1 GB. If enough files have the same write speed in a row, the write phase is ended early if 15 files haven't been written yet. After a short pause to allow any hidden caching to finish writing the data.
SSD Benchmark Results
After benchmarking a SSD, you can view the details of each set of files by clicking on the drive's icon.
From these results, you can surmise that the drive either does not have any built-in cache or it honored the request to disable cache writes. If you see higher bars at the start and then dropping significantly, the drive is caching it's writes and ignoring the cache bypass directive, typically to boost perceived performance..
Important
SSDs formatted with ZFS may not reflect true performance due to ZFS's built-in caching mechanisms that currently can't be fully bypassed.
The Harddrive Database has many drive images you can use if you don't like the default one or you can upload your own - such as there is no image in the HDDB available for your drive and you see the default gray drive icon. Clicking on a drive to view its details, you can click on the "Edit Drive" button to customize the image or add a text overlay with the capacity if it's not given in the image.
The best & easiest way to install applications is using UNRAID's Community Apps. Search for DiskSpeed and then click on the Install button for it.
Add a new Docker application with the following settings:
Name: DiskSpeed
Repository: jbartlett777/diskspeed
Network Type: Bridge
Use Tailscale: No
Console shell command: (No preference)
Privileged: Yes. This docker app scans the hardware subsystems to determine the store & drive configurations which requires Privileged set on
Web Port: 18888 (you can choose any)
Config Type: Path
Name: Local Storage
Container Path: /tmp/DiskSpeed
Host Path: /mnt/user/appdata/DiskSpeed/
Default Value: leave blank
Access Mode: Read/Write
Description: External storage location to store persistent data
Config Type: Path
Name: UNRAID
Container Path: /mnt/UNRAID
Host Path: /mnt
Default Value: leave blank
Access Mode: Read/Write
Description: Create this variable if you want to benchmark SSD drives
Config Type: Path
Name: unRAID ini Files
Container Path: /var/local/emhttp
Host Path: /var/local/emhttp
Default Value: /var/local/emhttp
Access Mode: Read/Only
Description: Location of unRAID ini files
docker create --name='DiskSpeed' --net='bridge' ---privileged=true -p '18888:8888/tcp' -v '/mnt/user/appdata/DiskSpeed/':'/tmp/DiskSpeed':'rw' -v '/mnt':'/mnt/UNRAID':'rw' -v '/var/local/emhttp':'/var/local/emhttp':'ro' 'jbartlett777/diskspeed'
Diskspeed Users can upload benchmarks to the HDDB to contribute to the performance statistics of the drives (in addition to being able to recover the benchmarks if they delete the app & reinstall later). You can see how different revisions of the same model perform against each other and it's also clear not all drives with the same model are created equal. You can also see this drive has the typical Seagate dip at the start of the drive - which is odd in its own right because drives are supposed to be faster at the start. Seagate's dip runs up to around 50 MB into the drive.
You can view the HDDB at https://www.strangejourney.net/hddb/