# [mrtg] Re: Whither the 95th percentile parser?

Henry Steinhauer H1STEINH at hewitt.com
Thu Dec 16 18:05:52 MET 1999

```Pete -

Thanks for your work.  You make it seem easy.

One question though;

Some of the vendors will use a different process of creating the 95th
value.

You appear to be creating the 95th value for in and the 95th for out  -
-  then seeing which one is larger and reporting that.

What some vendors are doing is comparing each interval and seeing which one
of the two is larger and keeping that value.
That way, they are jumping between in and out depending on which ever one
is higher at that measurement interval.

It tends to give a higher value for the 95th traffic value.  Hence they get
more money for the higher value.

But - for those of us with circuits that are not the same bandwidth in as
out, this still would not work.

Could I ask you to give me the snimp of code to:

1 -  give the higher of the two for a given date ?
I think that would be   \$level = \$in;
\$level = \$out if (\$out > \$level);
push (@highest,\$level);

2 - the 95th for in, out and Highest.
print "\$comment  usage for \$month_name \$year is    \$level   in is
\$high_in  out is  \$high_out   highest is \$highest_l  \n ";
where \$highest_l = @sort_high = sort {\$a <=> \$b } @highest;  \$highest_l =
\$sort_high[\$index];