Auslagerungsspeicher (Swap)
Übersicht
Der Auslagerungsspeicher wird zur Erweiterung des für die Ausführung von Programmen verfügbaren Speichers (RAM) verwendet.
- Ohne Auslagerungsspeicher werden bei Speichermangel die Anwendungen beendet, was zu einem Absturz des gesamten Systems führen kann.
- Wenn dem System Auslagerungsspeicher zur Verfügung steht, kann der Kernel weniger häufig genutzte Speicherseiten von inaktiven Anwendungen / Diensten auf die Festplatte schreiben ("auslagern"), wodurch mehr echter Arbeitsspeicher für aktive Anwendungen zur Verfügung steht.
Es gibt drei verschiedene Möglichkeiten, wie der Swap in Manjaro verwaltet werden kann:
- A swap partition
- A swap file
- zswap
This article tries to be as comprehensive as possible but even more information can be found in the articles linked below
Do I Need Swap
That is a question that cannot be answerd without having a look at your configuration and even with plenty of available memory, it is often used as a safety net or even sometimes due to specific application requirements so have a look at the following non-exhaustive list:
- If you use hibernation: yes, you need swap!
- If you have services that are not always active, but are still running all the time: yes, you need swap!
- If you have an application that allocates virtual memory directly for temporary storage instead of RAM: yes, you need swap!
- If you have an application that has a memory leak: yes, you need swap!
- If you have a server with 1TB of RAM that you're using as a desktop without applications allocating virtual memory or having memory leaks: No, you don't need swap!
How Much Swap do I Need
The amount of swap you need is highly variable based on your specific applications and workload. There is no universal formula on swap size without monitoring usage over a period of time. A reasonable place to start would be:
- For less then 4GB of physical memory (RAM), it's highly recommended that the swap space should, as a base minimum, be equal to the amount of RAM. Also, it's recommended that the swap space is maximum twice the amount of RAM depending upon the amount of disk space available for the system because of diminishing returns.
- For more modern systems (>4GB), your swap space should be at a minimum be equal to your physical memory (RAM) size if you use hibernation, otherwise you need a minimum of ROUNDUP(SQRT(RAM)) I.E. the square root of your RAM size rounded up to the next GB. The maximum, is again twice the amount of RAM, again because of diminishing returns.
- The only downside to having more swap space than you will actually use, is the disk space you will be reserving for it cannot be used for application or system data.
The "diminishing returns" means that if you need more swap space than twice your RAM size, you would be better off adding more RAM as Hard Disk Drive (HDD) access is about 10³ slower then RAM access, so something that would take 1 second, suddenly takes more then 15 minutes! And on a Solid State Drive (SSD) the same operation that took 1 second in RAM will still take about 1 minute on that SSD!
Taking into account all of the above, this brings us to the following table: (last 3 columns denote swap space)
RAM No hibernation With Hibernation Maximum
1GB 1GB 2GB 2GB
2GB 2GB 3GB 4GB
3GB 3GB 5GB 6GB
4GB 4GB 6GB 8GB
RAM No hibernation With Hibernation Maximum
5GB 2GB 7GB 10GB
6GB 2GB 8GB 12GB
8GB 3GB 11GB 16GB
12GB 3GB 15GB 24GB
16GB 4GB 20GB 32GB
24GB 5GB 29GB 48GB
32GB 6GB 38GB 64GB
64GB 8GB 72GB 128GB
128GB 11GB 139GB 256GB
256GB 16GB 272GB 512GB
512GB 23GB 535GB 1TB
1TB 32GB 1056GB 2TB
2TB 46GB 2094GB 4TB
4TB 64GB 4160GB 8TB
8TB 91GB 8283GB 16TB
Displaying Swap Information
The command swapon will display your current swap information. For example:
swapon NAME TYPE SIZE USED PRIO /dev/sda7 partition 20G 44.3M -2
The following script will:
- show whether zswap is active or not and if active, give zswap parameters if run with the Template:$ic command
- display a list of all applications / services that take up swap and how much they take up in descending order
#!/bin/bash #Check whether running as root if [ "$(whoami)" = 'root' ]; then dmesg | grep "zswap:" | grep --silent "load" if $? -eq 0 ; then # zswap is active echo "zswap information:" grep --recursive --color=none . /sys/kernel/debug/zswap/ read -n 1 -s -r -p "Press any key to continue" else echo "[warning] zwap not active. Continuing" fi else echo "[warning] Not running as root: skipping zswap info" fi for szFile in /proc/*/status ; do awk '/VmSwap|Name/{printf $2 "\t" $3}END{ print "" }' "$szFile" done | sort --key 2 --numeric --reverse | more
Using a Swap Partition
A swap partition is the traditional way of managing swap. In this scenario, a dedicated partition (or partitions) are created for holding swap.
Creating and Enabling a Swap Partition
To create a swap partition, you need enough unallocated disk space to create an additional partition. If you do not have enough space, skip to the next section #Using_a_swap_file. A swap partition can be created in any disk management / partition management tool and should be set as type linuxswap.
Once you have a swap partition you will need to initialize the swap partition with mkswap. For example, if your swap partition is /dev/sda3, you could use the command:
sudo mkswap /dev/sda3
Next we need to enable the swap partition with the swapon command. Following our example above this could be done with:
sudo swapon /dev/sda3
In order to ensure that the swap is enabled at boot we can add an entry to /etc/fstab. It is best to use the UUID instead of the device name for this purpose. You can add the line to fstab manually or using the command:
sudo bash -c "echo UUID=$(lsblk -no UUID /dev/sda3) none swap defaults 0 0 >> /etc/fstab"
If you would like to have more than one swap partition, simply repeat the steps above for any additional partitions.
Using a Swapfile
Using a swap partition has one major disadvantage: Changing the size of swap or adding swap requires repartitioning the disk. In current Linux kernels, it is possible to use a swap file instead of a dedicated partition and as from kernel 2.6 onwards there is no performance difference any more between the two.
Creating and Enabling a Static Swapfile
First create and intialize the file to hold the swap. For example, to create a 4GB swapfile, you could use the command:
sudo fallocate -l 4G /swapfile sudo mkswap /swapfile
Set the appropriate permissions on the file. It should be readable and writable only by root. This can be done with the command:
sudo chmod u=rw,go= /swapfile
Next we need to enable the swapfile with the swapon command. Following our example above this could be done with:
sudo swapon /swapfile
In order to ensure that the swap is enabled at boot we can add an entry to /etc/fstab. You can add the line to fstab manually or using the command:
sudo bash -c "echo /swapfile none swap defaults 0 0 >> /etc/fstab"
Swapfiles on BTRFS
As of kernel 5.0 and higher, swapfiles are supported on btfrs. They still require some special handling in addition to the above steps.
Prior to running the fallocate step above, you should run these commands:
sudo truncate -s 0 /swapfile sudo chattr +C /swapfile sudo btrfs property set /swapfile compression none
These commands create an empty swapfile, disable COW for that file and ensure that compression is disabled.
Swapfiles on ZFS
zfs doesn't support swapfiles, however you can achieve a similar benefit using a zvol as a swap volume.
Detailed instructions on how to accomplish this can be found in this ZoL guide.
Automated Swap Management with systemd-swap
It is possible to automatically manage the size of your swapfile using systemd-swap. It can create swap files dynamically based on memory needs. It is also an easy way to enable/manage zswap
Installing and Configuring systemd-swap
First, install and enable systemd-swap:
pamac install systemd-swap sudo systemctl enable systemd-swap.service
Next enable dynamic swap management by creating a file /etc/systemd/swap.conf.d/myswap.conf with the following parameters:
zswap_enabled=0 zram_enabled=0 swapfc_enabled=1
If you would prefer a single command to create this file you could use:
sudo bash -c 'echo -e "zswap_enabled=0\nzram_enabled=0\nswapfc_enabled=1" > /etc/systemd/swap.conf.d/myswap.conf'
Remove Other Swap
In order for systemd-swap to be effective you need to remove any traditional swap devices you have.
To display your current swap devices you can use the command swapon. For example, my test machine displays the following:
swapon NAME TYPE SIZE USED PRIO /swapfile file 4G 0B -2 /dev/sda3 partition 8.4G 0B -3
In this example there are two swap devices. A swapfile and a swap partition. To turn them off we can use the command swapoff.
sudo swapoff /swapfile sudo swapoff /dev/sda3
Next we need to remove them.
The swap partition will need to be deleted using your favorite partitioning tool. You can remove the swapfile by simply deleting it.
sudo rm /swapfile
We need to also remove any lines related to swap from /etc/fstab. Edit the file with an appropriate text editor and remove or comment out the lines for the swap space. In the example above, we would remove these lines:
UUID=c2430cf1-8ea9-4422-a5c5-5a38779194c3 swap swap defaults,noatime 0 2 /swapfile none swap defaults 0 0
Next reboot. After the reboot, when running swapon you should see something similiar to the following:
NAME TYPE SIZE USED PRIO /var/lib/systemd-swap/swapfc/1 file 512M 0B -2
systemd-swap will add and remove swapfiles as your memory usage dictates.
Using zswap with systemd-swap
zswap is compressed swap kept in RAM. zswap keeps the most frequently used pages in RAM and writes less frequently used pages to the swap space on disk. In many workloads, this will result in increased swap performance and less wear and tear on SSDs.
To enable zswap with systemd-swap, simply alter the instructions above and set Linux Kernel parameter Template:META Error
Tuning & Performance Considerations
Although swap seems like a great way to expand memory, excessive swap use will cause severe performance degradation, as mentioned before.
There are couple of parameters that can be used to tune swap utilization. These are swappiness and vfs_cache_pressure. To see your current settings for these you can use the following commands:
cat /proc/sys/vm/swappiness cat /proc/sys/vm/vfs_cache_pressure
swappiness controls how likely a page is to be transferred to swap. This value represents the percentage of the free memory before activating swap. The lower the value, the less swapping is used and the more memory pages are kept in physical memory where:
- 0 disables swap
- 60 is the default value which is ideal for a server running a lot of services
- 100 is very aggressive swapping.
For most computers the recommended value is 10: Theoretically, this means to only start swapping when RAM usage reaches around 90 percent.
vfs_cache_pressure is a percentage value that controls the tendency of the kernel to reclaim the memory which is used for caching of directory and inode objects. The default value is 100. Increasing this value will increase the rate in which these objects are removed from the RAM cache. Decreasing it will allow these objects to be cached in memory longer, consuming additional RAM over time. Depending on your specific workload, increasing or decreasing this value too far can have significant negative impacts on system performance. Experimentation is needed to find the appropriate balance and the default value is reasonable. In general, it is more common to optimize swappiness before experimenting with vfs_cache_pressure.
To set these values you can use the command sysctl. For example, to set the swappiness value to 10 you could use:
sudo sysctl vm.swappiness=10
There is no preset defined answer on the optimal values for these parameters. Experimentation is needed to find the optimum configuration for your specific hardware and workload.
See Also
- The Arch Wiki page on swap
- The Arch Wiki page on zswap
- The Linux kernel documentation on zswap
- The Linux kernel documentation on swappiness and vfs_cache_pressure
- The openSUSE guide to tuning memory
- The systemd-swap page
- The ZFS on Linux guide on Swap Volumes