ping command

he ping command is useful for determining the status of the network and various foreign hosts, tracking and isolating hardware and software problems, and testing, measuring, and managing networks

Some ping command options relevant to performance tuning are as follows:

-c

Specifies the number of packets. This option is useful when you get an IP trace log. You can capture a minimum of ping packets.

-s

Specifies the length of packets. You can use this option to check fragmentation and reassembly.

-f

Sends the packets at 10 ms intervals or immediately after each response. Only the root user can use this option.

If you need to load your network or systems, the -f option is convenient. For example, if you suspect that your problem is caused by a heavy load, load your environment intentionally to confirm your suspicion. Open several aixterm windows and run the ping -f command in each window. Your Ethernet utilization quickly gets to around 100 percent. The following is an example:

# date; ping -c 1000 -f 192.1.6.1 ; date
Thu Feb 12 10:51:00 CST 2004
PING 192.1.6.1 (192.1.6.1): 56 data bytes
.
--- 192.1.6.1 ping statistics ---
1000 packets transmitted, 1000 packets received, 0% packet loss
round-trip min/avg/max = 1/1/23 ms
Thu Feb 12 10:51:00 CST 2004

Note: The ping command can be very hard on a network and should be used with caution. Flood-pinging can only be performed by the root user.

In this example, 1000 packets were sent within 1 second. Be aware that this command uses IP and Internet Control Message Protocol (ICMP) protocol and therefore, no transport protocol (UDP/TCP) and application activities are involved. The measured data, such as round-trip time, does not reflect the total performance characteristics.

When you try to send a flood of packets to your destination, consider several points:

• Sending packets puts a load on your system.
• Use the netstat -i command to monitor the status of your network interface during the experiment. You may find that the system is dropping packets during a send by looking at the Oerrs output.
• You should also monitor other resources, such as mbufs and send/receive queue. It can be difficult to place a heavy load onto the destination system. Your system might be overloaded before the other system is.
• Consider the relativity of the results. If you want to monitor or test just one destination system, do the same experiment on some other systems for comparison, because your network or router might have a problem.