2017+1

>2017+1
>using swap space/pagefile
i shiggy diggy

Hibernate is nice.

>i shiggy diggy
Literally what did he mean by this?

>newfags

Some applications require it.
I think it was when I was doing Einstein@Home, if I didn't have any swap enabled even though RAM was not being fully utilized the applications would crash or reset.

In short, have some swap. You don't need much and if the system isn't using it then it isn't using it. It won't adversely affect your performance just sitting there.

Swap can be useful, but not using GPT in year 2002+2^2^2 is a sin

...

Just create a swapfile on your RAMdisk

Either way that is just swap you created.

>2018
>not using swap in a zramdisk to magically have more vm than you actually have without swapping to disk

Comes handy when running the science botnet

swap is always logical you dumb fuck.

lurk moar

>I have no idea what virtual memory is: The Post

No it's not.

I have 1GB of RAM

>It won't adversely affect your performance just sitting there.
A program misbehaves and starts eating up memory.

>you don't have swap
The OOM killer kills the bad process.

>you have swap
System starts swapping furiously and locks up. The only way to fix the problem is to pull the cord.

You are what is killing this website

Only if you're planning on having more than 3 other primary partitions. On a system with just / /home and swap, it really doesn't matter.

>2018 - 4hours
>not knowing what shiggy diggy means
ISHYGDT

>ISHYGDT
>not ISHYGDDT
DUN GOOF'D

I was hoping no one would notice!

Yes, you have no idea what virtual memory is.

There are very specific conditions required for swap to be advantageous (assuming Linux's MM which I am familiar with, no idea what wincock does). Having more memory is always faster than having less memory plus swap, so if you can afford it, it's always better to buy more memory/servers than add swap.

Basically, in order for swap to be useful, first of all the maximum simultaneous process memory and page cache has to exceed the amount of RAM you have. Otherwise, the kernel will eagerly swap process memory to disk when you have enough RAM and actually slow your computer down.

Second, there must be enough idle process memory to be worth swapping out. If all of you processes actively use their memory, the kernel will just thrash and your computer will slow the fuck down (remember slow Windows around 2000? That's basically all due to swap thrashing.) You would be much better off just buying the memory that you need instead if using swap. Furthermore, there is no advantage to making more swap than this swappable set. In theory there is no harm in making more swap than you need if you actually need swap, except if a process starts using up lots of memory your computer will lock up instead of killing the process automatically, so you're just fucking yourself over because you're choosing between killing one misbehaving process and killing all process when you force shut down the machine. So you're better off just creating the minimum amount of swap, if you actually need it.

Note that this means that swap is good if you run a program that leaks memory and that you don't restart often (like Firefox kek) since the kernel will eventually just start swapping out leaked memory which will never get used.

Fpbp and /thread

Oom killer doesn't kill the bad process, it kills a random process.

Wouldn't it be faster to just put in a 128mb stick and just a massive page file on 2 asus hyper m.2s with 4x Samsung 1tb SM961 VROC? I mean hitting the 128 gigabit read limit on VROC is a lot higher IOPs than ram. You could have 16GB/s reads and potentually 8TB of swap space. Every bench for a ramdisk only ever gets 5-8GB/s reads, and around 7-10GB/s writes. Even if you were to have some sort of retarded 4.5GHz ram, this solution is still higher IOPs, but I guess latency might be a little less, but the only reason latency is even important is because lower latency means faster RAM polling.

If one process eats up 30GB and the rest of the processes are under 1GB, the OOM killer will kill that process.

The OOM killer only kills "randomly" if your system just doesn't have enough RAM, not if you have a rogue process.

>he doesn't have at least 64GB of RAM

bought 16GB DIMMs at 75 dollars last year.

>Otherwise, the kernel will eagerly swap process memory to disk when you have enough RAM and actually slow your computer down.

> I really don't understand how virtual memory works: part 2

Just stop, you dumb motherfucker. Clearly you have just enough surface knowledge about swap but no more and you're hanging yourself on your ignorance.

>Fuck, I don't know what he's talking about. Better call him stupid without providing any argument. I wrote a thesis involving compressed caching in the Linux MM, dear user. The concept of virtual memory isn't even relevant when discussing swap performance, which is how I know you know nothing; you keep throwing around the phrase virtual memory since it's the only thing you have a basic grasp on.