This test is described in more detail in the following links, which we refer to:
Test plan: Experiment F Plan - Flow Subsystem Burst Throughput
Test result: Master: Experiment F - Flow Subsystem Burst Throughput
The difference is that we use the IXIA tester to simulate openflow switches, total 100 switches was set and 100000 flows for each switch, the results show below.
Flow_mode Rate
Flow_mode rate(P/S) Single mode(ONOS) | Flow_mode rate(P/S) Three mode(ONOS1) | Flow_mode rate(P/S) Three mode(ONOS2) | Flow_mode rate(P/S) Three mode(ONOS3) | |
|---|---|---|---|---|
| First | 10760 | 13112 | 12569 | 8941 |
| Second | 11414 | 8383 | 7865 | 11244 |
| Third | 11892 | 8558 | 7376 | 8477 |
| Fourth | 12029 | 7983 | 7790 | 9028 |
| Fifth | 12174 | 13387 | 13087 | 7666 |
| Sixth | 11319 | 9126 | 8905 | 8335 |
| Seventh | 11225 | 14442 | 13546 | 7906 |
| Eighth | 12585 | 10446 | 9748 | 12318 |
| Ninth | 12065 | 10456 | 9878 | 8406 |
| Tenth | 10413 | 9638 | 9237 | 13580 |
| Avg | 11587 | 10553 | 10000 | 9590 |
| Std | 644 | 2188 | 2165 | 1939 |
As the histograms show, in the case of three nodes, flow_mod is three times the case of a single node, but the standard deviation is much higher than single node.