| Cheshire Cat Computing http://steveshipway.org/forum/ |
|
| MRTG/ROUTERS data sampling resolution and aging http://steveshipway.org/forum/viewtopic.php?f=2&t=717 |
Page 1 of 1 |
| Author: | cracraft [ Tue Sep 13, 2005 5:58 am ] |
| Post subject: | MRTG/ROUTERS data sampling resolution and aging |
Hi -- in the daily view, the data points represent a 5 minute sampling period since our scripts are waking up that often, collecting the data from the system, and sending it over for processing by mrtg. Correct me if I am wrong on the above. What happens to the sampling period as the data ages to weekly monthly and yearly? Thank you (again), Stuart |
|
| Author: | stevesh [ Tue Sep 13, 2005 11:56 am ] |
| Post subject: | |
(This is more of a MRTG question, and is dealt with in more detail in the MRTG documentation) In RRD, each RRA is defined by a number of data points and a consolidation factor. If your data sample is 5-minute, then your Daily Average line will be of an RRA that is 1 data point, Average -- IE, the sampled data itself. The weekly graph is 6 data points, Average (and the purple Max line is 6dp,Max). This means that what you are seeing is the average over 6x5min, IE, the 30-min average. Similarly, Monthly gives a 2hr average and Yearly gives a 1day average. Where it getsa bit more complex is the Max lines. These show the Max *datapoint* value, so the yearly Max line actually shows the maximum 5-min average over the period of a day, since a data point is 5-min. If your sample Interval is 1-min then things are similar, but with 5 times as many data points (the 6-hr graph is the 1dp graph). People with a mathematical bent will have now realised why the 95th Percentile calculations are necessarily inaccurate and become more inaccurate as the timeframe widens. This is again explained mathematically in the routers2 documentation. So, in summary, the sampling period is still 5-min, but RRD collects the required number of data points and averages them to get a value which is stored. The XFF value, which defaults to 50%, is the proportion of datapoints in the set which must exist for this calculation to be considered valid - else an 'unknown' is stored. |
|
| Page 1 of 1 | All times are UTC + 12 hours [ DST ] |
| Powered by phpBB® Forum Software © phpBB Group http://www.phpbb.com/ |
|