Chapter 5. Operating system considerations 363
6. If you are using striped DMS tablespaces (multiple containers) on ESS,
ensure the extent size is a multiple of the rank’s stripe width (normally
32 KBytes).
7. I/O servers are used by database agents for prefetch I/O or asynchronous I/O
by utilities (for example, backup or restore). Having more than required does
not impose a large load on the system. Configure NUM_IOSERVERS equal to the
number of physical disks used by the database.
8. Asynchronous page cleaners flush dirty pages from the buffer pools to
maintain a certain amount of free spaces. The number of page cleaners can
have a significant impact on performance. Increase the number of page
cleaners (NUM_IOCLEANERS) if the database is heavily update-oriented or has a
large number of buffer pool pages. This is typically set to the number of CPUs
on the server.
For additional information about RAID and IBM ESS exploitation with DB2, refer
to the IBM Redbooks Database Performance on AIX, SG24-5511, and IBM ESS
and IBM DB2 UDB Working Together, SG24-6262.
5.2.4 Monitoring and problem determination tools
This section provides a brief description of UNIX commands available for
monitoring and tuning the performance of the operating system.
Table 5-4 lists tools that are commonly used to monitor various resources of
interest (this is not a comprehensive list).
Table 5-4 Suggested tool usage
This section covers the following topics:
򐂰 Mapping filesystems to physical disks
Note: In most cases, these commands require root or other privileges only
held by the UNIX sysadmin, and
not by the DBA. Therefore, the DBA will most
likely have to coordinate with the UNIX syadmin to monitor the system
resources of interest.
Performance problem Suggested tools
High CPU utilization vmstat, sar, nmon,
Not enough memory vmstat, vmtune, nmon, svmon
Paging and/or swapping vmstat, vmtune, nmon, lsps, Memory Visualizer
Disk or filesystem iostat, sar, nmon, filemon
364 DB2 UDB ESE V8 Performance Guide for High Performance OLTP and BI
򐂰 filemon
򐂰 iostat
򐂰 lsps
򐂰 nmon
򐂰 ps
򐂰 System activity recorder (sar)
򐂰 svmon
򐂰 vmstat
򐂰 vmtune
A brief overview of each tool is provided, along with sample output.
Mapping filesystems to physical disks
To determine disk utilization and placement relating to specific DB2 objects, it is
necessary to map filesystems to the physical disks on which they reside.
Figure 5-11 shows a sample filesystem
6
associated with a logical volume.
Figure 5-11 Relationship between filesystems and disks
This subsection provides a brief overview of the steps involved in determining the
relationship using a number of UNIX commands, as follows:
6
Adding a filesystem writes essential data structures in the target volume, and it becomes usable
after this filesystem is mounted
filesysystem
hdisk hdisk
pdisk pdisk
df
device specific commands
lslv -l
Logical Volume
Chapter 5. Operating system considerations 365
1. Identify the logical volume associated with a file system by using the df
command.
2. Identify the “hdisks” associated with the logical volume by using the lslv -l
command.
3. Identify the “pdisks” associated with a given “hdisk” by using the ssaraid
command.
For example, in Figure 5-12, we show how to identify the physical disks
associated with filesystem /sys.
Figure 5-12 Mapping a filesystem to disks
The df /sys command in Figure 5-12 displays the filesystem mounted as /sys.
The Filesystem column shows the device on which the filesystem resides. For
local filesystems, these are found in /dev. In this case, the device is the logical
volume named systemsw.
The lslv -l systemsw command
7
in Figure 5-12 lists a single hdisk, named
hdisk4, as being associated with the filesystem as reported in the PV column.
7
The -l option limits the report to the four columns shown.

Get DB2 UDB ESE V8 non-DPF Performance Guide for High Performance OLTP and BI now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.