Firefox and Thunar, if I remember correctly (in Openbox, no less). I deduced it to be swap related because when everything was done on the SSD as opposed to the microSD, there were no problems of the sort.
It is my understanding that disabling swap entirely would make it worse, because then it would have no recourse once the memory capacity has been filled.
Except for Raspberry Pi4, the bus for USB, ethernet, storage, is all shared, and typically the bottleneck.
i.e. whether you have a HDD or SSD the throughput is equally poor in my experience.
You might have had a very slow SD card however, and that would make a difference.
Not having a swap, means that it will swap out pages, but i have found the performance improves, as it's not treating very slow disk as memory, and instead just uses keeps currently accessed pages
in memory if that makes sense
[automerge]1589781978[/automerge]
Firefox and Thunar, if I remember correctly (in Openbox, no less). I deduced it to be swap related because when everything was done on the SSD as opposed to the microSD, there were no problems of the sort.
It is my understanding that disabling swap entirely would make it worse, because then it would have no recourse once the memory capacity has been filled.
Looks like you are correct that the SD card is a
very slow write.
Just did a quick and dirty test (on a raspberry pi in use - so not a perfect test scenario).
Creating a zero-ed 1GB file (with a small 1024 block size) resulted in:
SD Card (class 10 ext3) - around 5-6 MB/s
HDD (USB 2 in RAID - so maybe slightly slower than a native HDD. ext4) - 21 MB/s
USB stick (USB 2 and vfat
) - 26 MB/s
USB stick (USB 2 and exfat
fuse file system)- 4 MB/s
Lot of variables there I know. I have a ext3 or 4 formatted USB somewhere I will report back if it makes any significant difference in speed.
Obviously block size will make a big difference too