Using sed commands on an output to find an average - bash

Question

My goal is to compute the average ping times of a website over the first 10 counts, average over the second 10 counts and average over the last 10 counts and print them in one go using bash scripting.
So far, I have come up with the following script:
ping -c 30 google.com > pings.txt
sed 's/\=/= /g' pings.txt > formatted_op.txt
sed -n '2, 11p' formatted_op.txt > 1.txt
sed -n '12, 22p' formatted_op.txt > 2.txt
sed -n '22, 32p' formatted_op.txt > 3.txt
awk '{print $10}' 1.txt | awk '{ sum += $1 } END {print sum/10}' > 1_avg.txt
awk '{print $10}' 2.txt | awk '{ sum += $1 } END {print sum/10}' > 2_avg.txt
awk '{print $10}' 3.txt | awk '{ sum += $1 } END {print sum/10}' > 3_avg.txt
cat 1_avg.txt 2_avg.txt 3_avg.txt > final_avg.txt
cat final_avg.txt
rm 1.txt 2.txt 3.txt 1_avg.txt 2_avg.txt 3_avg.txt formatted_op.txt pings.txt
However, I want to be able to do this without creating any temporary files. How do I do so?
I have also tried doing it using pipes as follows:
ping -c 30 google.com | sed 's/\=/= /g' | sed -n '2, 11p' | awk '{print $10}'| awk '{ sum += $1 } END {print sum/10}'
but this only computes the average of the first 10 pings and using ping 3 times would give different results each time.

Answer 1

This can all be done with awk pretty easily. Assuming your ping output looks like this:
64 bytes from yyz08s09-in-f110.1e100.net (172.217.1.110): icmp_seq=27 ttl=59 time=0.636 ms
64 bytes from yyz08s09-in-f110.1e100.net (172.217.1.110): icmp_seq=28 ttl=59 time=0.638 ms
64 bytes from yyz08s09-in-f110.1e100.net (172.217.1.110): icmp_seq=29 ttl=59 time=0.658 ms
64 bytes from yyz08s09-in-f110.1e100.net (172.217.1.110): icmp_seq=30 ttl=59 time=0.666 ms
This will do the trick:
ping -c 30 google.com | \
awk '
{
split($8,a,"=");
if(NR > 1 && NR < 12) {
round1+=a[2]
} else if (NR < 22) {
round2+=a[2]
} else if (NR < 32) {
round3+=a[2]
}
}
END {
print round1/10" "round2/10" "round3/10
}
'
We use the NR variable to check which line of output is being processed and then increment the appropriate variable. (The value is gotten by splitting the time field on the equals sign.)

Answer 2

You don't have to make it that complicated. Do something like
for i in {1..10}
do
ping -c 30 google.com | tail -n1 |
awk -v count="$i" -v FS="/" '{print "Count", count,"average : ",$5}'
done
Sample Output
Count 1 average : 78.484
Count 2 average : 74.473
Count 3 average : 76.971
Count 4 average : 78.789
Count 5 average : 103.609
Count 6 average : 105.754
Count 7 average : 98.969
Count 8 average : 99.009
Count 9 average : 86.186
Count 10 average : 86.521

Answer 3

A pure Bash solution is not possible, because Bash has no support for floating point arithmetic. You need at least bc.
In order to calculate an average it is not necessary to collect and store all values. You can calculate the average incrementally. See here or here for the mathematics or read Volume 2 of Donald Knuth's "The Art of Computer Programming". The formula as a Bash function might look like this:
avg ()
{
local a=$1 # average of the values already seen in the past
local x=$2 # new sample
local n=$3 # sequence number
bc -l <<<"$a + ($x - $a) / $n"
}
A single ping time can be collected by the following function.
pingtime ()
{
ping -c1 ${host:-localhost} |
sed -n 's/.*time=\(.*\) .*/\1/p'
}
With the above functions the calculation of n average values with m samples for each average can be done with two for loops.
n=3
m=10
for i in $(seq $n); do
a=
for j in $(seq $m); do
t=$(pingtime)
if [ -z "$a" ]; then
a=$t
else
a=$(avg $a $t $j)
fi
done
LC_NUMERIC=C printf '%2g\n' $a
done
In the outer loop the average a is first cleared. In the inner loop the ping time is measured. If the average is empty it is initialized with the first sample. Otherwise the average is calculated. And the result is reported in the outer loop.
bc and printf behave differently depending on the localization. If necessary a specific behavior can be forced by setting LC_NUMERIC.
If you wish you can add a sleep 1 in front of the ping, to measure only one sample per second.

Answer 4

You can also try my solution
#!/bin/bash
pings=10
for i in {1..10};
do
ping -c ${pings} google.pl | grep -o 'time=.*' | sed 's/time=\(.*\) ms/\1/g' | awk -v iteration="$i" -v pings="$pings" '{sum+=$1+$2+$3} END {print "Iteration " iteration " Average time " sum/pings}'
done