Juniper J-Series Administration Manual page 291

UDP and TCP probe types require that the remote server be configured as an RPM
receiver so that it generates responses to the probes.
RPM Tests
Each probed target is monitored over the course of a test. A test represents a collection
of probes, sent out at regular intervals, as defined in the configuration. Statistics are
then returned for each test. Because a test is a collection of probes that have been
monitored over some amount of time, test statistics such as standard deviation and
jitter can be calculated and included with the average probe statistics.
Probe and Test Intervals
Within a test, RPM probes are sent at regular intervals, configured in seconds. When
the total number of probes has been sent and the corresponding responses received,
the test is complete. You can manually set the probe interval for each test to control
how the RPM test is conducted.
After all the probes for a particular test have been sent, the test begins again. The
time between tests is the test interval. You can manually set the test interval to tune
RPM performance.
Jitter Measurement with Hardware Timestamping
Jitter is the difference in relative transit time between two consecutive probes.
You can timestamp the following RPM probes to improve the measurement of latency
or jitter:
NOTE: The Services Router supports hardware timestamping of UDP ping and UDP
ping timestamp RPM probes only if the destination port is UDP-ECHO (port 7).
Timestamping takes place during the forwarding process of the Services Router
originating the probe (the RPM client), but not on the remote router that is the target
of the probe (the RPM server).
The supported encapsulations on a Services Router for timestamping are Ethernet
including VLAN, synchronous PPP, and Frame Relay. The only logical interface
supported is an
ICMP ping
ICMP ping timestamp
UDP ping
UDP ping timestamp
services interface.
lt
Chapter 14: Configuring RPM Probes
RPM Overview
269