Unix System Performance

This chapter deals exclusively with data pulse points for Unix systems. Information contained in this chapter includes:

"Processor Performance."

"Memory Performance."

"Disk Performance."

"Summary of Unix Pulse Points."

The following Unix pulse points are provided by John Herberg of Lund Performance Solutions.

Each section also contains a handy table that readily identifies at which points your data is normal, questionable, or unacceptable.

Processor Performance

This section discusses five graphs offered by Performance Gallery Gold that focus on HP-UX processor performance:

"CPU Busy Percentage."

"CPU High Priority Busy Percentage."

"Real Time Processing Percentage."

"Run Queue Average."

"System Processing Percentage."

CPU Busy Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
CPU Busy %	less than 60	60 to 85	greater than 85

The CPU Busy % pulse point shows the percentage of time the CPU spent executing the following activities instead of being in a pause or idle state:

Processing user and system process code.

Managing main memory.

Scheduling and dispatching processes (interrupts).

Processing context switches and overhead (external device activity).

This information is reported in the Performance Gallery Gold CPU Detail (UX) line graph (see the data element titled "cpu>cpu-busy%").

Figure 10.1 Example CPU Detail (UX) line graph: cpu> cpu-busy%

CPU High Priority Busy Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
CPU High Priority Busy %	less than 60	60 to 85	greater than 85

The CPU High Priority Busy % pulse point shows the percentage of time the CPU spent executing high-priority system and user processes, interrupts, and overhead.

This information is reported in the Performance Gallery Gold CPU Utilization (UX) line graph (see the data element titled "CPU high pri").

Figure 10.2 Example CPU Utilization (UX) line graph: CPU high pri

Real Time Processing Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Real Time Processing %	less than 5	5 to 10	greater than 10

The Real Time Processing % pulse point shows the percentage of time the CPU spent serving online, interactive sessions ("real time" user processes). These processes run at a fixed high- priority status.

This information is reported in the Performance Gallery Gold CPU Utilization by Category (UX) stacked area graph (see the data element titled "Real").

Figure 10.3 Example CPU Utilization by Category (UX) stacked area graph: Real

Run Queue Average

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Run Queue Average	less than 5	5 to 10	greater than 10

The Run Queue Average pulse point shows the average number of executable processes that waited for the CPU during a collection interval.

This information is reported in the Performance Gallery Gold CPU Queue Detail (UX) line graph (see the data elements titled "Queue 5 min average").

Figure 10.4 Example CPU Queue Detail (UX) line graph: Queue 5 min average

System Processing Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
System Processing %	less than 10	10 to 20	greater than 20

The System Processing % pulse point shows the percentage of time the CPU spent executing system calls or operating in kernel mode.

This information is reported in the Performance Gallery Gold CPU Utilization by Workload (UX) subchart of system processes.

Figure 10.5 Example CPU Utilization by Workload (UX) subchart (sysprocs)

Memory Performance

This section discusses three graphs offered by Performance Gallery Gold that focus on HP-UX memory performance:

"Memory Used Percentage."

"Page Outs per Second."

"Deactivations per Second."

Memory Used Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Memory Used %	less than 80	80 to 90	greater than 90

The Memory Used % pulse point shows the average percentage of main memory used during the collection interval.

This information is reported in the Performance Gallery Gold Memory Used % / Deactivations Rate (UX) line graph (see the data element titled "Memory used percent").

Figure 10.6 Example Memory Used % / Deactivations Rate line graph: Memory used %

Page Outs per Second

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Page Outs/second	less than 5	5 to 10	greater than 10

The Page Outs/second pulse point shows the number of instances per second that a page out occurred during the collection interval. A page out is performed to move the least-needed pages from memory by writing them to swap space or to the file system. A page out occurs when physical memory becomes scarce.

This information is reported in the Performance Gallery Gold Memory Rd Hit % / Page Fault Rate (UX) line graph (see the data element titled "Page outs/sec").

Figure 10.7 Example Memory Rd Hit % / Page Fault Rate (UX) line graph: Page outs/sec

Deactivations per Second

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Deactivations/second	less than 2	2 to 5	greater than 5

The Deactivations/second (swap outs/second) pulse point shows the number of processes swapped out of memory to disk in order to satisfy extreme memory shortages.

This information is reported in the Performance Gallery Gold Memory Used % / Deactivations Rate (UX) line graph (see the data element titled "Deactivations/sec").

Figure 10.8 Example Memory Used % / Deactivations Rate line graph: Deactivations/sec

Disk Performance

This section discusses three graphs offered by Performance Gallery Gold that focus on HP-UX disk performance:

"Disk Queue Length."

"Read Hit Percentage."

"Disk I/O Rate."

Disk Queue Length

Performance Indicator	Performance Ranges
	Normal	Problematic	Unacceptable
Disk Queue Length	less than 1	1 to 3	greater than 3

The Disk Queue Length pulse point shows the average number of processes in the request queue for a particular disk drive.

This information is reported in the Performance Gallery Gold Disk I/O Queue Length (UX) line graph.

Figure 10.9 Example Disk I/O Queue Length (UX) line graph

Read Hit Percentage

Performance Indicator	Performance Ranges
	Normal	Problematic	Unnacceptable
Read Hit %	greater than 90	90-80	less than 80

The Read Hit % pulse point shows the percentage of time that requests for information were satisfied in main memory. The Read Hit % value should be above 90. A percentage less than 90 could indicate a data locality problem or a shortage of memory.

This information is recorded in the Performance Gallery Gold Memory Rd Hit % / Page Fault Rate (UX) line graph (see the data element titled "Read hit percent").

Figure 10.10 Example Memory Rd Hit % / Page Fault Rate (UX) line graph: Read hit percent

Disk I/O Rate

Performance Indicator	Performance Ranges
	Normal	Problematic	Unnacceptable
Disk I/O’s per second	less than 40	40 to 60	greater than 60

The Disk I/O Rate pulse point shows the number of disk I/O (reads and writes to disk) per second.

This information is reported in the Performance Gallery Gold Disk I/O by Drive (UX) graph. This stacked area graph plots the disk I/O rate recorded for each disk drive.

To view the disk I/O rate for a particular disk drive, do the following:

Open the Disk I/O by Drive (UX) chart.

From the Chart menu, select Modify Chart.

From the Modify Chart dialog box, select the disk drive from the Subchart drop-down menu.

Click OK.

Figure 10.11 Example Disk I/O Rate (UX) subchart, c0t5d0 disk

Summary of Unix Pulse Points

For your convenience, all of the Unix pulse points have been summarized in the following table.

Performance Indicator	Performance Ranges
	Normal	Problematic	Unnacceptable
Processor Performance
CPU Busy %	less than 60	60 to 85	greater than 85
CPU High Pri Busy %	less than 60	60 to 85	greater than 85
Real Time Processing %	less than 5	5 to 10	greater than 10
Run Queue Average	less than 5	5 to 10	greater than 10
System Processing %	less than 10	10 to 20	greater than 20
Memory Performance
Memory Used %	less than 80	80 to 90	greater than 90
Page Outs/second	less than 5	5 to 10	greater than 10
Deactivations/second	less than 2	2 to 5	greater than 5
Disk Performance
Disk Queue Length	less than 1	1 to 3	greater than 3
Read Hit %	greater than 90	90-80	less than 80
Disk I/O Rate	less than 40	40 to 60	greater than 60