README for nmon for Linux version 11d

The postings on this site solely reflect the personal views of the authors and do not necessarily represent the views, positions, strategies or opinions of IBM or IBM management.
This new nmon for Linux version 11 has the following new features:

1. Improved user interface
2. User Defined Disk Groups
3. Bug fixes
4. Better filesystem stats
5. Ignore thresholds can be set
6. Change directory on startup
7. Can handle more the 256 disks
8. 2.6 kernel CPU Wait for I/O works
9. 2.6 kernel virtual memoary stats added

1) Improved user interface

The user online interface has been improved to the same style as the AIX nmon version.
This means some curses boxes and lines are used and better headlines for the different statistics.
I think it looks more professional.

2) User Defined Disk Groups

This feature allows you to define the groups for disks. Check the nmon documentations for the details as this has been available for the nmon for AIX version for some time.

One urgent request was tracking the disks connected to a particualr HDA - this can now be done.
Also you often find that the disk stats are double what you might expect as the /proc files with the data include botht he whole disk (as in /dev/sda) and the partitions (as in /dev/sda1 and 2 etc). User Defined Disk Groups can be used to remove unwanted details. See the nmon FAQ for more details of why nmon does not remove disks.
Use the -g option and name a file of a file containing the User Defined Disk Groups definitions.

3) Bug fixes

A couple of bugs came up including the failure to run and collect data in to a file on larger systems.
Sorry about that but it has been removed with version 11.
The new nmon Wiki allows quick beta testing versions of new releases means this should not happen again.
Online type g to see the stats.

4) Better filesystem stats

The filesystems stats (the j option online) now work better and report filesystem that are mounted but not in the /etc/fstab.
There seems to be s growing number pseudo filesystems these days these are more accurately spotted and highlighted.

5) Ignore thresholds can be set

Now you can set the threshold at which processes and disks are ignores in the dot mode (.) or data collection to file.
The default is 0.1% but you can lower it with, for example, -I 0.0001
This becomes important with ever faster machines.

6) Change directory on startup

This is useful for running nmon from a cron. The -m <directory> option saves you having to use a schell scrit to start nmon from a cron file.

7) Can handle more the 256 disks

Some users have gone beyond 256 disks. You can now use the -d <disks> options to go higher then the default number of disks.

8) 2.6 kernel CPU Wait for I/O works

This is available with the Linux 2.6 kernel only using undocumented fields in the /proc/stat file.
With this version and previous kernel versions CPU wait for I/O is always reported as zero.
Previous version sof nmon reported CPU nice and CPU wait for I/O !

9) 2.6 kernel virtual memoary stats added

This is available with the Linux 2.6 kernel only using an undocumented file called /proc/vmstat
To activate this online use V (capital v).
This is switched on by default in the save to file mode.
All fields in this file are displayed and saved - now we have to work out what they all mean !!!
The nmon Analyser Excel spread sheet 3.2 onwards handles data OK but the graphs might need tweaking.

If you want other things, then let me know at http://www.ibm.com/eserver/pseries/community/ and take the performance tools link.
Partiularly, if you know where the data can be found in the /proc filesystem or other files or system calls.

Thanks Nigel Griffiths, IBM UK