| 1 |
On 8/20/22 10:22 PM, William Kenworthy wrote: |
| 2 |
> What are you measuring the speed with - hdparm or rsync or ? |
| 3 |
> |
| 4 |
> hdparm is best for profiling just the harddisk (tallks to the interface |
| 5 |
> and can bypass the cache depending on settings, rsync/cp/?? usually have |
| 6 |
> the whole OS storage chain including encryption affecting throughput. |
| 7 |
|
| 8 |
How you measure performance is a complicated thing. There is the raw |
| 9 |
device speed verses the speed of the system under normal load while |
| 10 |
interacting with the drive. |
| 11 |
|
| 12 |
At $WORK, we are more concerned about throughput of the drive in our day |
| 13 |
to day use case than drive's raw capacity. |
| 14 |
|
| 15 |
> Encryption itself can be highly variable depending on what you use and |
| 16 |
> usually though not always includes compression before encryption. |
| 17 |
|
| 18 |
Compression can be a very tricky thing. There's the time to decompress |
| 19 |
and compress the data as it's read and written (respectively). Then |
| 20 |
there's the throughput of data to the drive and through the drive to the |
| 21 |
media. If you're dealing with text that can get a high compression |
| 22 |
ratio with little CPU overhead, then there's a good chance that you will |
| 23 |
get more data into / out of the drive faster if it's compressed than at |
| 24 |
the same bit speed decompressed. |
| 25 |
|
| 26 |
To whit I enabled compression on my ZFS pools a long time ago and never |
| 27 |
looked back. |
| 28 |
|
| 29 |
> There are tools you can use to isolate where the slowdown occurs. |
| 30 |
> atop is another one that may help. |
| 31 |
|
| 32 |
Yep. |
| 33 |
|
| 34 |
> [test using a USB3 shingled drive on a 32 it arm system] |
| 35 |
|
| 36 |
Is that an Odroid XU4 system? If so, why 32-bit vs 64-bit? -- Or am I |
| 37 |
mistaken in thinking the Odroid XU4 is 64-bit? |
| 38 |
|
| 39 |
> xu4 ~ # hdparm -Tt /dev/sda |
| 40 |
> /dev/sda: |
| 41 |
> Timing cached reads: 1596 MB in 2.00 seconds = 798.93 MB/sec |
| 42 |
> Timing buffered disk reads: 526 MB in 3.01 seconds = 174.99 MB/sec |
| 43 |
> xu4 ~ # |
| 44 |
|
| 45 |
If that is an Odroid XU4, then I strongly suspect that /dev/sda is |
| 46 |
passing through a USB interface. So ... I'd take those numbers with a |
| 47 |
grain of salt. -- If the system is working for you, then by all means |
| 48 |
more power to you. |
| 49 |
|
| 50 |
I found that my Odroid XU4 was /almost/ fast enough to be my daily |
| 51 |
driver. But the fan would kick in for some things and I didn't care for |
| 52 |
the noise of the stock fan. I've not yet compared contemporary |
| 53 |
Raspberry Pi 4 or other comparable systems. |
| 54 |
|
| 55 |
|
| 56 |
|
| 57 |
-- |
| 58 |
Grant. . . . |
| 59 |
unix || die |
Archives