# Computation Method and Aggregation

The computation methods used to calculate aggregate traffic from source data are described in the table here.

Computation Method | Description |
---|---|

Max | For each hour of 24 hours, the maximum of the sample values within that hour is used. |

Average | For each hour of 24 hours, the samples within that hour are averaged. If there are N samples within an hour, the result is the sum of the all the sample values divided by N. |

95th percentile | For each hour of 24 hours, the 95th percentile value
of the samples within that hour is used. The 95th percentile is computed
from the equation: Average + 1.645 * stddev, where Average refers
to the average of the samples in that hour and stddev is the standard
deviation. Another way of thinking about this is that 95 percent of
the sample values lie below the calculated value. The corresponding
equation for the 90th percentile is: |

**Roll-up / Aggregation Defined**

The intention of traffic aggregation is to summarize the network traffic data that is generated across several days of collection. Selected days or a range of days can be specified to be included in the aggregated result. For example, one can aggregate all Mondays in the month of January, weekdays only, or aggregate all the days in an entire month, months, or year.

For example, in the aggregated data for a particular interface, there are 24 values representing the 24 hours. For that particular interface, all sample points in hour 1 (In bps and Out bps values) across all the days are grouped together, all sample points in hour 2 across all the days are grouped together, and so on. The 95th percentile for hour 1 then takes into consideration as sample points, all the sample points for that interface during hour 1 on all the selected days.

**Note**

Aggregate directories have the same format as the daily directories and can also be aggregated. This might be useful when performing the offline traffic processing incrementally as more traffic data becomes available. However it is up to you to remember exactly which days of traffic were aggregated - usually by supplying a meaningful name the directory.