locked
MAP Server Consolidation 95th percentile calculation HELP RRS feed

  • Question

  • Since Server consolidation considers 95th percentile value 

    I was just reading through on how 95th percentile is calculated and per the article http://www.netcraftsmen.com/95th-percentile-calculation/  :-

    • Collect all the data samples for a period of time (commonly a day, a week, or a month).
    • Sort the data set by value from highest to lowest and discard the highest 5% of the sorted samples.
    • The next highest sample is the 95th percentile value for the data set.

    Wouldn't that give wrong data at times, taking example of a machine's 95th percentile of %CPU utilization & Memory Utilization.

    Consider an example where I ran the Server Consolidation report in MAP toolkit for 60minutes with sample interval of 5minutes. 

    • So I'll have 12 samples (CPU %) like - 2,2,2,2,2,2,2,2,2,2,40,60,51
    • Sorting them - 2,2,2,2,2,2,2,2,2,2,40,51,60
    • Discarding 5% of sorted samples i.e. discarding 1 top sample - 60(Discarded)
    • So next highest sample which is 95thpercentile value = 51

    Now in this case for all the time server utilization was 2% and only for 3 intervals(40,60,51) it peaked and may have dropped down right after it reported the last value of 60%.
    Which basically would mean for majority of the time server utilization was just 2% but since for that 95th interval it peaked high, it'll consider 51% as the value for consolidation.

    Wouldn't that be wrong to consider 51% utilization for calculations for this machine where I should be actually considering just 2% for this machine.

    TIA



    • Edited by Yankee30 Thursday, December 3, 2015 10:13 PM
    Thursday, December 3, 2015 10:11 PM