2019-03-22 update

content/blog/testing-hard-drive-ssd-performance.md

@@ -0,0 +1,125 @@
+---
+title: "Testing hard-drive or SSD performance on Fedora"
+date: 2019-03-22T00:00:00+02:00
+opensource: 
+- Fedora
+---
+
+If your Linux system appears to be slow, it might be an issue with your disks,
+either hard drive or SSD. Hopefully, with a few commands you can get an idea
+of the performances of your disks.
+
+First, you will have to install `hdparm` using `yum` or `dnf`:
+
+```sh
+sudo yum install hdparm
+```
+
+Then, you will need to know which disk to test (your system might have multiple
+disks). In this example, let's pretend we find our `/` partition to be slow.
+
+The `df` command will show you on which disk is hosted the `/` filesystem:
+
+```raw
+# df -h /
+Filesystem                Size  Used Avail Use% Mounted on
+/dev/mapper/storage-root   10G  2.6G  7.5G  26% /
+```
+
+In this example, the `/` partition is on a logical volume (`root`) that is in
+the `storage` volume group. If you suspect another partition, replace `/` with
+the mountpoint of the other partition (`/var` for instance).
+
+Let's find on which disk(s) is the volume group `storage` with the `pvdisplay`
+command:
+
+```raw
+$ sudo pvdisplay
+  --- Physical volume ---
+  PV Name               /dev/sda2
+  VG Name               storage
+  PV Size               <557.88 GiB / not usable 4.00 MiB
+  Allocatable           yes
+  PE Size               4.00 MiB
+  Total PE              142816
+  Free PE               89056
+  Allocated PE          53760
+  PV UUID               9FT0O7-zs52-U11S-ROL1-ixVW-tgL9-9fVKXC
+```
+
+In the previous output, there is one physical volume (`/dev/sda2`) in the
+`storage` volume group. So we need to measure the performances of `/dev/sda`.
+
+The `hdparm -tT` command performs timing reads of the disk, with and without
+the buffer cache of the Linux kernel.
+
+```raw
+$ sudo hdparm -tT /dev/sda
+
+/dev/sda:
+ Timing cached reads:   25732 MB in  1.99 seconds = 12926.07 MB/sec
+ Timing buffered disk reads: 556 MB in  3.00 seconds = 185.23 MB/sec
+```
+
+According the man page, the `-t` parameter is used to:
+
+> display the speed of reading directly
+> from the Linux buffer cache without disk access.  This measurement is
+> essentially an indication of the throughput of the processor, cache, and
+> memory of the system under test.
+
+According the man page, the `-T` parameter is used to:
+
+> display the speed of reading
+> through the buffer cache to the disk without any prior caching of data.
+> This measurement is an indication of how fast the drive can sustain
+> sequential data reads under Linux, without any filesystem overhead.
+
+You can test the write speed with the `dd` command. This command will create a
+file and write to it. The `conv=fdatasync` is used to actually write the content
+to disk by flushing the buffers.
+
+```sh
+sudo dd if=/dev/zero of=/output conv=fdatasync bs=256k count=1k; rm -f /output
+```
+
+Make sure you write the file in the filesystem you want to test. For instance,
+if you want to test the write speed on `/var`, replace `/output` with
+`/var/output`.
+
+```raw
+$ sudo dd if=/dev/zero of=/output conv=fdatasync bs=256k count=128; rm -f /output
+128+0 records in
+128+0 records out
+33554432 bytes (34 MB) copied, 0.0259041 s, 1.3 GB/s
+
+$ sudo dd if=/dev/zero of=/output conv=fdatasync bs=256k count=512; rm -f /output
+512+0 records in
+512+0 records out
+134217728 bytes (134 MB) copied, 0.110927 s, 1.2 GB/s
+
+$ sudo dd if=/dev/zero of=/output conv=fdatasync bs=256k count=1k; rm -f /output
+1024+0 records in
+1024+0 records out
+268435456 bytes (268 MB) copied, 0.32287 s, 831 MB/s
+
+$ sudo dd if=/dev/zero of=/output conv=fdatasync bs=256k count=8k; rm -f /output
+8192+0 records in
+8192+0 records out
+2147483648 bytes (2.1 GB) copied, 11.2527 s, 191 MB/s
+```
+
+Note: your write speed might be faster than your read speed. How can it be so?
+Well, if your server has an high-end hardware disk controller, it might have
+dedicated hardware buffers in addition to a small battery to retain data in case
+of a power loss. In this case, if the file is small enough to fit the hardware
+buffer of the disk controller, you are actually mesuring the bandwidth of your disk
+controller.
+
+To measure the write speed of your disk, you have to write a file **much larger**
+than the hardware buffers of your disk controller.
+
+In the previous examples, the hardware buffers of my server seems to be around
+128 MB (the performances start to drop after that).
+
+So the actual write speed of my disk is around 190 MB/s.