| 180 | === test three: 80% upload, 20% download, 1MB mean file size === |
| 181 | |
| 182 | Same environment as test 2, but 80% of the operations are uploads. |
| 183 | |
| 184 | {{{ |
| 185 | clients: 150kBps down, 680kBps up, .14 fps down, .67 fps up |
| 186 | tahoecs1: 62% CPU, 11Mbps out, 2.9Mbps in, load avg .85 |
| 187 | tahoecs2: 57% CPU, 10Mbps out, 4Mbps in, load avg .76 |
| 188 | tahoebs4: 16% CPU, 700kBps out, 5.4Mbps in, load avg 0.4ish |
| 189 | tahoebs5: 21%, 870kBps out, 5.1Mbps in, load avg about 0.35 |
| 190 | }}} |
| 191 | |
| 192 | Overall throughput is about half of the download case. Either uploading files |
| 193 | or modifying the dirnodes looks to be more expensive than downloading. The |
| 194 | CPU usage on the web servers was lower, suggesting that the expense might be |
| 195 | in round trips rather than actual computation. |
| 196 | |