Hardware Capabilities
In this group of examples, the performance assessment tool measures the speed of specific computers—individually or in a group.
Optimal Sustained Receive Rate
What is the maximum rate at which a specific computer can receive a stream of messages?
To answer this question, run rvperfs
on the receiving computer. Then run rvperfm
in automatic mode on another computer; select a message size that reflects the messages that your application will send when deployed.
receiver>rvperfs
sender>rvperfm -auto -size 1000
After rvperfm
experiments with its parameters, the final run indicates the values that yield the optimal receive rate for the receiving computer under prevailing network conditions. (However, you must validate this measurement; see below.)
A similar test with several receivers determines the optimal rate of the slowest receiver.
To validate an optimal receive rate, check that it is strictly less than the maximum transfer rate from rvperfm
to rvd
(see below).
• | If rvperfm on the sending computer has successfully transferred a run of messages at a rate strictly greater than the optimal receive rate, then that receive rate is valid. |
• | If the measured receive rate is approximately equal to the maximum transfer rate, it might be because some limitation on the sending host is causing an artificially low result for the receive test. |
To obtain the sender’s maximum transfer rate, run rvperfm
in automatic mode on the sending computer, without any rvperfs
processes to receive the messages; use the same message size as in the receive test.
sender> rvperfm -auto -size 1000
After rvperfm
experiments with its parameters, the final run indicates the values that yield the maximum transfer rate to the daemon. This result is not a useful measure of network performance; its only legitimate use is to validate measurements of receive rates.
Fixed Receive Rate
Can all computers on this network receive 2000-byte messages at a sustained rate of approximately 5 batches per second, with 10 messages per batch?
To answer this question, run rvperfs
on each of the receiving computers. Then run rvperfm
in single mode on another computer.
receiver1>rvperfs
receiver2>rvperfs
...
receiver42>rvperfs
sender>rvperfm -size 2000 -batch 10 -interval .2
The run report indicates whether the receivers keep pace with the sender under prevailing network conditions.