Closed karapirinc closed 1 year ago
I think the confusion stems from the fact, that we are using "moving average" (with very small window). While it may seem "incorrect" in the whole execution period, it gives better indication if there was a spike in the processing time. See the example below: Processing Time | (moving) Average processing time | (Total) Average processing time |
---|---|---|
2 | ||
4 | 3 | 3 |
6 | 5 | 4 |
3 | 4 | 4 |
3 | 3 | 4 |
3 | 3 | 4 |
6 | 5 | 4 |
7 | 6 | 4 |
8 | 7 | 5 |
3 | 5 | 5 |
6 | 5 | 5 |
7 | 6 | 5 |
33 | 20 | 7 |
6333 | 3176 | 459 |
6 | 1591 | 429 |
3 | 797 | 402 |
2 | 400 | 379 |
2 | 201 | 358 |
77 | 139 | 343 |
8 | 73 | 326 |
4 | 39 | 311 |
7 | 23 | 297 |
2 | 12 | 284 |
444 | 228 | 291 |
6 | 117 | 279 |
3 | 60 | 269 |
3 | 32 | 259 |
3 | 17 | 250 |
3 | 10 | 241 |
3 | 7 | 233 |
Average processing time calculation does not present actual average time.
tigase.stats.StatisticsInvocationHandler.Statistics#updateExecutionTime
Suppose that, there are 3 executions with 3, 5 and 10ms. Expected average processing time : (3 + 5 + 10) / 3 = 18 / 3 = 6ms Actual average processing time : ( (3 + 5) / 2 + 10 ) / 2 = 7ms
Over time gap between actual and expected times increases. Actual average processing time closer to the last value.