I liked the server I set up the other day. 512 gigs of ram, 1 gig swap.
We’re using maybe 100gigs currently.
Why even bother having swap at that point?
because swap does other things than “extending” your physical ram
I think the question is: if a person is going to make such a tiny swap, why even use swap?
Such a small swap is unlikely to save a system from memory problems and it’s does not seem likely to make a noticeable difference in performance when it’s only able to swap out small amounts of memory.
Why wouldn’t one just put in larger ZRAM or a larger Swap with a reduced swapiness?
If I have a raspberry pi with 1 GB ram, I don’t think a 2 MB swap is worth bothering with.
Swap is not intended to save a system from memory problems. And absolutely not fit for that purpose.
And if you are using that memory, it will make a noticeable difference in performance.
But for it to have any impact at all, you need the default (high) swapiness. If you reduce it, that partition will become useless.
Anyway, I don’t recommend you put any swap at your PI’s SD card.
I wouldn’t put swap on an SD card, no. Even if it had an NVME, it seems like putting up at least a double-digit percent would be more effective than 1%.
Also, since 6.1, swap has been a lot better, with MGLRU. ChromeOS gets away with paltry amounts of RAM due to swapping. So classic overcommitting seems fine as long as you don’t run into situations where more RAM is active at once than is available by hardware.
4 GB are 4 GB, it doesn’t matter how much memory you have at total.
The only usage of swap where you should even look at your RAM size is for on-disk sleeping. And most people don’t do this anymore.
Because it’s not what what you think it is and doesn’t do what you think it does. https://chrisdown.name/2018/01/02/in-defence-of-swap.html
Disabling swap does not prevent disk I/O from becoming a problem under memory contention, it simply shifts the disk I/O thrashing from anonymous pages to file pages
While the rest of that post matches my understanding of swap (I still think 1GB is next to useless in this case), that summarized point perplexes me.
What non-special file(s) does the kernel write to and read from, and how does it know how much space to use?
That’d be over 1TB with zram on
This comment is how I learned about zram. Just one question, when is it used?
Fedora enables it on all systems with <4GB (maybe a bit less I forget). It’s trivial to enable/disable it and see if its helpful for your usage.
Yep. I just looked it up.
I assumed it was like windows page file but it’s ramdisk.
Depends on config, ArchWiki recommends optmizing some sysctl values to take advantage of it
it generally starts kicking in after >60% RAM usage even with this config
Me watching my kernel shove 16gb worth of in use virtual memory into zswap in a desperate attempt to run 20 heavy applications on 8gb of physical memory
use a swap file