Operation Manual

TIP: Cylinder Numbers
Note, that different partitioning tools may start counting the cylinders of a
partition with 0 or with 1. When calculating the number of cylinders, you should
always use the difference between the last and the rst cylinder number and
add one.
Using swap
Swap is used to extend the available physical memory. It is then possible to use more
memory than physical RAM available. The memory management system of kernels
before 2.4.10 needed swap as a safety measure. Then, if you did not have twice the size
of your RAM in swap, the performance of the system suffered. These limitations no
longer exist.
Linux uses a page called “Least Recently Used” (LRU) to select pages that might be
moved from memory to disk. Therefore, running applications have more memory
available and caching works more smoothly.
If an application tries to allocate the maximum allowed memory, problems with swap
can arise. There are three major scenarios to look at:
System with no swap
The application gets the maximum allowed memory. All caches are freed, and thus
all other running applications are slowed. After a few minutes, the kernel's out-of-
memory kill mechanism activates and kills the process.
System with medіum sized swap (128 MB–512 MB)
At rst, the system slows like a system without swap. After all physical RAM has
been allocated, swap space is used as well. At this point, the system becomes very
slow and it becomes impossible to run commands from remote. Depending on the
speed of the hard disks that run the swap space, the system stays in this condition
for about 10 to 15 minutes until the out-of-memory kill mechanism resolves the
issue. Note that you will need a certain amount of swap if the computer needs to
perform a “suspend to disk”. In that case, the swap size should be large enough to
contain the necessary data from memory (512 MB–1GB).
Advanced Disk Setup 45