Monitoring Network Performance

To determine if the system is experiencing a network performance bottleneck network performance data needs to be gathered to determine the maximum throughput the system is capible with the the installed network hardware and software. Performance measurements tools sends packets from and to a machine measuring the bits per second, bps, or millions of bits per second, Mbps, that travel from end to end. After this has been accomplished, the workload can be monitored. Interactive monitoring tools track of the number of network packets arriving and departing from the system at the current instance of time, thus measuring the network workload.

The command netstat is used to report a variety of network information including the amount of kernel memory utilization. Kernel memory buffers are called mbufs. Mbufs are used to provide fast transfer of incoming network data to the kernel and the lack of enough mbufs can cause network throughtput to plummet. Below is a sample output from the netstat -m command from the IRIX operating system. If there are any requests for memory denied listed in the output the system may need more mbufs allocated. Allocating more mbufs is most safely performed by adding additional memory, however on some Unix operating systems the amount of memory dedicated to mbufs is configurable.

# netstat -m
346/431 mbufs in use:
170 mbufs allocated to data
3 mbufs allocated to packet headers
66 mbufs allocated to socket structures
95 mbufs allocated to protocol control blocks
5 mbufs allocated to routing table entries
4 mbufs allocated to socket names and addresses
2 mbufs allocated to interface addresses
1 mbufs allocated to internet multicast options
20/22 mapped pages in use
88 Kbytes allocated to network (87% in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines
.
.
.

• Solaris 2.5 or SunOS 5.5
• SGI IRIX 5.3
• HP HP-UX 10.0
• Linux

Terms used: packet, network, Mbps, buffer, kernel.