IBM TotalStorage DS6000 Series: Performance Monitoring and Tuning

Chapter 7. Open systems servers - UNIX 251

dev = /dev/305glv

vfs = jfs2

log = /dev/jfs2log01

mount = true

options = rw

account = false

{CCF-part2:root}/tmp/perf/scripts -> lsfs -q /interdiskfs

Name Nodename Mount Pt VFS Size Options Auto Accounting

/dev/305glv -- /interdiskfs jfs2 639893504 rw yes no

(lv size: 639893504, fs size: 639893504, block size: 4096, sparse files: yes, inline log: no,

From Example 7-37 on page 250 we notice:

򐂰 We specified a specific jfs2log for our file system. With several large, busy file systems

using the same jfslog, the jfslog can get very busy. We have demonstrated how to create

and use a separate jfs2log for a specific file system.

򐂰 The default file system block size of 4096 K was assigned to this file system when it was

created. This size is recommended.

7.9 Operating system tuning

Operating system tuning recommendations are outside the scope of this book. However,

because tuning at this level can improve I/O, we will overview some areas of interest. We will,

of course, cover AIX in more detail than other UNIX operating systems because we know AIX

better. File system buffer tuning should be done carefully and incrementally.

When tuning the operating system, do one thing at a time; verify I/O improvement every step

of the way. Have a clear understand of your current system settings before making changes to

the operating system.

Response time and throughput trade-offs exist that affect overall performance. Generally a

multi-user system will want to ensure good response time for users, while a system that runs

only batch jobs should be tuned for maximum throughput. The appropriate tuning parameters

are dependent upon the nature of the application, the number of CPUs, the amount of cache

in the DS6000, the write rate, and other factors. We suggest that you tune these values for

maximum disk throughput with reasonable response time for users.

7.9.1 AIX operating system tuning (JFS and JFS2)

Beginning with AIX 5.2, kernel tuning has been redesigned. The vmtune has been replaced by

the ioo (for I/O related parameters) and vmo (for pure VMM parameters) commands. Similarly,

the schedtune command has been replaced with the schedo command. In AIX 5.3 the vmtune

and schedtune commands no longer exist.

The new commands are part of the bos.perf.tune fileset in AIX and they all use the same

syntax and command options to manipulate files in the /etc/tunables directory. Also, starting

with AIX 5.2, SMIT provides full support for these commands.

A full discussion of AIX tuning can be found in the AIX 5L Version 5.3 Performance

Management Guide which can be found by first selecting the AIX documentation link, then

selecting the Performance management and tuning link, at:

http://publib.boulder.ibm.com/infocenter/pseries/index.jsp