Chapter 4. Tuning the operating system 115
Draft Document for Review May 4, 2007 11:35 am 4285ch04.fm
cache and the cache of the disk subsystem can no longer accommodate the amount or size
of a read or write request, the physical disk spindles have to work. Consider the following
example. A disk device is able to handle 200 I/Os per second. You have an application that
performs 4kB write requests at random locations on the file systems so streaming or request
merging is not an option. The maximum throughput of the specified disk subsystem is now:
I/Os per second of physical disk * request size = maximum throughput
Hence the example above results in:
200 * 4kB = 800kB
Since the 800kB is a physical maximum, the only possibility to improve performance in this
case is to either add more spindles or physical disks or to cause the application to write larger
I/Os. Databases such as DB2 can be configured to use larger request sizes that will in most
cases improve disk throughput.
For additional, in-depth coverage of the available IBM storage solutions, see:
򐂰IBM System Storage Solutions Handbook, SG24-5250
򐂰Introduction to Storage Area Networks, SG24-5470
Guidelines for setting up partitions
A partition is a contiguous set of blocks on a drive that are treated as if they were independent
disks. The default installation of today’s Enterprise Linux distributions use rather flexible
partitioning layouts by creating one or more logical volumes.
There is a great deal of debate in Linux circles about the optimal disk partition. A single root
partition method may lead to problems in the future if you decide to redefine the partitions
because of new or updated requirements. On the other hand, too many partitions can lead to
a file system management problem. During the installation process, Linux distributions enable
you to create a multipartition layout.
There are benefits to running Linux on a multipartitioned or even logical volume disk:
򐂰Improved security with finer granularity on file system attributes
For example, the /var and /tmp partitions are created with attributes that permit very easy
access for all users and processes on the system and are susceptible to malicious access.
By isolating these partitions to separate disks, you can reduce the impact on system
availability if these partitions have to be rebuilt or recovered.
򐂰Improved data integrity, as loss of data with a disk crash would be isolated to the affected
partition
For example, if there is no RAID implementation on the system (software or hardware) and
the server suffers a disk crash, only the partitions on that bad disk would have to be
repaired or recovered.
򐂰New installations and upgrades can be done without affecting other more static partitions.
For example, if the /home file system has not been separated to another partition, it will be
overwritten during an OS upgrade, losing all user files stored on it.
򐂰More efficient backup process
Partition layouts must be designed with backup tools in mind. It is important to understand
whether backup tools operate on partition boundaries or on a more granular level like file
systems.
Table4-4 lists some of the partitions that you may want to consider separating out from root to
provide more flexibility and better performance in your environment.