S3 file hashing scheme for linear distribution

gistfile1.txt

#!/bin/sh
# Get crazy throughput if you do this (>5000 puts/sec)
#
# What some customers have done is something like this: say this is the filename:
# 
#     mybucket/17476/26466/16.png
#
# Then to get the hash do something that seems silly but works great (pretty
# much any two characters from the hash will work):
#
#     echo '/17476/26466/' | openssl dgst -md5 -hex | cut -c22-23
#
# So this key ends up being:
#
#     mybucket/6e/17476/26466/16.png
#

echo $filename | cut -d/ -f1 | openssl dgst -md5 -hex | cut -c22-23