| Parameter |
Guideline Values
|
SAP Transaction
|
Unix Command
|
Comments |
|
CPU Management
|
|
|
|
|
| Utilization |
Utilization < 65% idle > 20% |
OS06 |
vmstast -S 5 9 |
High CPU utilization indicates that the database machine
needs more and faster cpus or a upgrade and for the application server we
might need to distribute loads across the other application server |
| Processor Activity |
Utilization < 65% idle > 20% |
OS06 |
sar -M, top, vmstat -S 5 9 |
System/Memory/Process Monitor - Use this to determine what
processes are consuming the most memory and cpu |
| Load Average |
>3 processes in queue = problem |
OS06 or ST06 |
--- |
OS06 shows an average of more than 3 processes waiting in
front of the CPU, then its an alarm to increase the CPU’s. |
| Process Information |
High values in column b is an indicator of a I/O bound
system.
High values in column r is an indication that the CPU is
a problem. |
---- |
vmstat -S 5 9 |
The three columns under procs show the number of processes
in the run queue (r), the number of blocked processes (b), and the number of
swapped runnable or short sleeper processes (w). High values of ‘r’ is an
indication that CPU performance is a problem. |
|
Memory Management
|
|
|
|
|
| % Free |
How much Physical memory is actually available to use. |
OS06 |
--- |
Check from Goto à Current data à Previous hoursà Memory. In
a small hardware configuration, CPU bottlenecks are apparent when physical
memory free is less than 10 MB. |
| Page In /Page Out |
Minimize page
activity. |
|
|
|
|
Swap Memory |
|
--- |
|
|
| Data Buffer Hit Ratio |
|
|
|
SAP
recommends that you maintain a
data
buffer quality of at least 97% on a production R/3 System. |
|
Redo Log Buffer Allocation
Fault Rate |
Fault rate is
= ratio of redo allocation retries to the total redo entries. |
|
|
This fault rate at 0% is ideal. |
|
Data Dictionary cache Hit Ratio |
> 90% |
|
|
%. |
|
SQL
Area getratio / pinratio |
|
|
|
% and pinratio %. |
|
Recursive Call Ratio |
Ratio of recursive calls to user calls < 10% |
|
|
If the number of
Recursive calls is greater than the number of User calls, then
you should start a detailed examination. Recursive Calls are a result
of low DD cache. |
|
Database
Management |
|
|
|
|
|
Full Table Scans |
|
|
|
|
|
Un-indexed Tables |
All tables should have at least a primary
index |
|
|
|
|
Disk
Management |
|
|
|
|
|
High Disk Activity |
|
|
|
|
|
Long Disk Queues |
queue < 2-4 |
--- |
|
|
|
DBWR Processes |
to the # of controllers |
|
--- |
|
|
Balanced Disk I/O |
Would like to even distribution of load -
|
|
|
|
