Performance analysis commands can execute on one node, but analyze data collected from other nodes. Performance Manager's performance analysis commands are scripts that detect performance problems and offer corrective advice in four areas: CPU, memory, network, and disk I/O.

CPU Commands

CPU Analysis

This script determines how efficiently a computer's CPU is being used. High idle time during a heavy load indicates an I/O bottleneck. High system time under a heavy load indicates excessive overhead. If inefficiency is discovered, other scripts can reveal the cause; try the Virtual Memory, Swapping, and Device I/O scripts.

Load Average

This script determines a computer's load average for the last minute, last 5 minutes, and last 15 minutes. The load average is the number of jobs in the run queue. An acceptable load average is 3 to 7 jobs for a large system, 1 to 2 jobs for a workstation. This script also reports if a computer is consumed by a small number of user processes, and lists the top CPU-using processes.

Memory Commands

Buffer Cache

This script determines if a computer's buffer cache is too large or too small. A too-small cache causes excessive I/O. A too-large cache causes excessive paging and swapping.

Excessive Paging

This script determines if there is excessive paging on a computer by checking the number of free pages, paged out pages, and page faults. Excessive paging can be caused by a new process trying to allocate pages, or by active virtual memory being too large relative to active real memory.

Excessive Swapping

This script displays virtual memory and swap space usage and detects excessive usage.

Memory Shortage

This script determines if a computer has a memory shortage. If there is much swapping during paging, and runnable processes are swapped out while the free list increases, lack of memory could cause desperation swapping (also called "thrashing") to occur.

Virtual Memory

This script determines if a computer has virtual memory problems. This script displays swap configurations and the number of free pages, and compares the amounts of physical and virtual memory.

Network Commands

Gateway Errors

This script determines if a computer has excessive gateway errors by looking at the number of bad checksum fields for IP, ICMP, TCP and UDP. Gateway errors should be fewer than one hundredth of a percent of the total number of packets received.

Network Errors

This script determines if a network node (a computer in a network) has exceeded the acceptable number of network output errors and collisions. This script examines the length of the send queue for all connections, and displays the number of output errors, input errors, and collisions, as well as the number of in and out packets.

Packet Retransmissions

This script determines if a node has excessive network packet retransmissions by looking at the number of retransmissions and bad xids. (Bad xids are acks that return an xid different from the one sent.) Packet retransmissions should be fewer than 1% of the total number of client NFS calls. Retransmissions increase when you are working with network hardware or all your computers boot at the same time.

Disk I/O Commands

Excessive Transactions

This script displays the transactions per second (tps) and total transactions on each device and reports excessive activity.

File System Analysis

This script determines if there are sufficient inode and file table entries to support the number of system processes. If inode and open file usage are more than 80%, increase the system parameter to make the usage less than 80%.