Importance of Swap Partition

The swap partition serves a couple of purposes.

• It serves as 'backup' RAM. That is, should your computer run out of RAM, it will use the swap area as a temporary source of more memory. More specifically, it will 'swap' unused items from the RAM into the swap area in order to leave spare space for the applications that need it at that instant. This is not ideal as the data transfer rate to the hard drive is significantly lower than that to your normal RAM. In practice this means its much slower to retrieve information from the swap area.

• It is used when the computer hibernates. Hibernation involves taking an image of the RAM in its current state (like an ISO represents an image), and saves it to the swap area. It then reloads this image when the computer restarts.

• To be most useful, the swap area should be at least (RAM * 1.5) although more is recommended. For example, on my system with 3gb of RAM, I have a swap area of 7.2gb.

This is very close to the same as this question about the “right” size for a swap partition. Much of the same information from my answer there applies - basically, if you want to hibernate you generally want your swap space to be at least as big as your RAM, and other than that a round number like 1 or 2 GB is easily sufficient. Because swap is so much slower than RAM, if you're filling up multiple gigabytes of swap your computer has almost certainly become unusably slow.

There's also no real need for a swap partition - swap files (available on the mainstream linux filesystems) give the same performance and make it trivially easy to add more swap space if you decide you haven't got enough.

here a very deep information about swap

some people say the double of your ram but personally i recommend this :

swap = 1.5 X Total Ram

Example :

if you have 2Gb of ram -> swap = 1.5 x 2 = 3

P.D : Ubuntu Desktop uses Swap to Hibernate (PC off, no power needed, program states saved). If Hibernation is important to you, have more swap space then ram + swap overflow.

The free command can tell you how much swap you are using. For example on this machine:

$ free -m
             total       used       free     buffers     cached
Mem:          1947       1863         84         312        758
-/+ buffers/cache:        792       1154
Swap:         4000          3       3997

shows me that I have 2GB (1947m) RAM and that the system has used most of it. However, 312m is used for I/O buffers and the remainder (758m) the system has decided to fill with disk cache.

The disk cache is interesting because it is using fast memory instead of slow disk for its contents. The contents could be gotten from disk, but they are kept around in case they are needed. This also means that there is 758m of memory that can be reclaimed in an instant if necessary because the system knows it can find that data on the disk instead.

That is why there is a second line showing that if there were no buffering and cache, I'd have half my RAM (1154m) available for use.

The third line shows that I have an overly large swap partition (it was there and wasn't doing anything) of which a whopping 3m have been used. This is stuff that the kernel really doesn't expect to have to use anytime soon so it was stuck out on the "back porch".

While free gives you the snapshot now, vmstat can give you a running picture:

$ vmstat 10
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
 0  0   3588  86236 316524 769132    0    0    14    13  126   81  4  1 95  0
 0  0   3588  83872 316532 770512    0    0     0    20  264 1229  3  1 96  0

There's a lot of information there, but of interest is that there is no swap-in (si) or swap-out (so) traffic. Which means I'm not using the swap at all over the last 10 seconds.

I have a laptop with a slow hard drive, but relatively high RAM (8GB). On this setup, I find a swap drive to be counterproductive for the following reasons:

1. Getting applications back out of swap is slow, comparable to loading the application from scratch. I would rather have an environment that responds quickly, even if that means I sometimes have to close and reload applications, than an environment which sometimes stalls on me unexpectedly.

2. If I wanted the ability to hibernate, that would mean writing a lot of data to disk. That would be a constant drain on battery if it was done persistently (the default behaviour of swappiness 60), or if it was only done at hibernation time (by reducing swapiness), then it would take a long time to actually hibernate. I found suspending to be RAM instead of suspending to disk to be quite satisfactory. It could stay suspended for a couple of days, and it would unsuspend much more quickly than a full hibernation.

3. Because the hard drive is slow, I prefer to keep 1GB of RAM reserved for disk cache, effectively exchanging memory for speed. This keeps my machine snappy, but it does mean I can use only 7GB of my 8GB of RAM for applications. The software I use to achieve that is a fork of earlyoom which kills old browser tabs when 88% of the RAM has been used. I occasionally have to reload old browser tabs which have been killed off.

When would this advice not apply?

• If you have a fast hard disk drive, or an SSD, and have fewer concerns about battery drain, then the disadvantages I outlined above will not apply. (Although early SSDs might have issues with wear, I believe any SSDs sold since 2018 should be fine.)

• If your machine has very little RAM, or you like to use multiple memory-hungry applications at the same time, then you may need to use swap to get your work done, regardless of the disadvantages. This was the primary use-case for swap, before memory started to become cheaper.

• If you really need hibernation (perhaps you want to suspend your laptop, and re-open it after a few days without charging) then you need a swap drive (recommended at least as large as your RAM).

Caveat:

• Some of the software that we leave running for a long time (especially the OS and desktop environment) will load code into memory which will never actually be used again. These pages could be moved to swap and will never need to be pulled out again, thus avoiding the disadvantages I mentioned earlier. For this reason, I sometimes create a small 512MB or 1GB swap file, and let the unused pages of memory get swapped out. This just leaves me with a little more RAM for applications. (Needed: A technique for measuring swap churn, to help tune the size of this swap file.)

• Swap area is the part of Hard Disk space used to support limited space RAM memory. RAM is limited in size so, some applications need more RAM than the available RAM space, In that case this Swap Space also called as Virtual Memory used to support RAM.
• It is slower in speed compared to RAM's speed. When we run a large application then the least recently used part of that application in swapped to Swap Area on the Hard Drive, it is swapped back from Swap Area when it is needed. This gives Operating System a feel of just having more RAM than actually it is.
• It is a dedicated partition on Hard Disk created while installing Operating System.
• It is good if create Swap Area of Double the size of RAM.

• You can check used and available Swap Area using following command:

  $ cat /proc/swaps 
  Filename                Type        Size    Used    Priority
  /dev/sda10              partition   7812092  16      -1