CSIRO Radiophysics Laboratory Network Performance Monitoring Methodology

The Curtin University perf Tools

The information included here has been extracted from the README file that is included in the software distribution.

Perf attempts to check network performance issues. It uses a particular version of the well-known ping utility to send ICMP echo-request packets to various router interfaces and measures the round trip times and packet loss. The version of ping used comes from a NetBSD distribution. Packets are sent in a similar way to a Cisco style ping. A packet is sent, and on receipt of the reply, another packet is sent. If no reply is received a further packet is sent after an interval of one second.

Perf sends a user defined number of packets to a selected IP address at random intervals during each hour. The rationale in using this algorithm is that with the possibility of numerous utilities checking network performance, everyone is not checking each link at exactly the same time. The latter would be more of a check on the checking process itself than the link itself!

A timeout value, measured in milliseconds is attached to each entity being monitored. If packets are received beyond this value they are not counted as successful returns but as late arrivals.

The size of packets was determined by looking at normal usage patterns in Internet Protocol traffic. The 64, 576 and 1536 byte packets were regularly seen. As this code is running under Solaris 2.5 the maximum data size is 1472 bytes for ICMP so this highest value is not monitored here.