11/24   Have a stream of data (latency numbers for webpage requests).
        Need the running median, approximate ok.  How do I get this
        without storing all the data?
        \_ Are you the dude asking about 'variability' earlier?
        \_ Make 'bins', say 0-10ms, 11-20ms, 21-30ms, ... and store how many
           samples fall into each bin.  You can then find the median within
           <binsize> accuracy and it will use <max-latency>/<binsize> ints.
           Latencies greater than <max-latency> can be put in their own special
           bin but don't forget them when finding the median.
           \_ Thanks. -op
           Now, assuming you have this setup and know the true average, how do
           you find the standard deviation?
        \_ Try R: http://www.r-project.org