conductor/beacon

Fun with the NIST randomness beacon

#Introduction#

DevOps is about managing production environments at large scale -- deep stacks, hundreds of servers, highly distributed systems. Such systems tend towards decay and chaos. It's just a fact of the job: Something is probably going wrong at any given moment in a large enough installation.

This exercise tests your ability to distinguish the signal from the noise. First, you will organize a random input into meaningful output. Second, you'll report on that output using software.

#Prerequisites#

Your answer will be written in python, and must meet the following requirements:

• The python component should run on python v3.2.3+, unless you import a public module that requires v2.7.x. If you do so, you'll be expected to justify that decision. Python programs that rely on pre-2.7 syntax will not be accepted.

• It must run on Ubuntu 12.04 (aka "precise"), and shouldn't require any unusual binaries or new software to run.

If you're a vagrant user, you can exactly replicate the box on which we'll test your answer by using this base box and then running apt-get install -y python3.

#The Randomness Beacon#

The National Institute of Standards and Technology broadcasts a signed, timestamped random bit sequence once a minute, every minute, and has since Unix epoch time 1378395540 (09/05/2013 11:39 a.m.). It's called the Randomness Beacon, and you're going to analyze it.

##Goal 1: Sampling Chaos##

The Randomness Beacon publishes a RESTful API, through which you can retrieve any given minute's beacon value as 128 hexadecimal characters, as well as a variety of other authentication data. Your first goal is to summarize the beacon's emissions over arbitrary amounts of time.

This repository contains the basic structure of a python module. Note that the module set-up has been accomplished for you, and python setup.py install, issued inside a clone of this repository, should work. Within the bin/ directory, you'll find a summarize-beacon stub.

Rewrite the summarize-beacon stub to be a full-fledged python script, relying, if you'd like, on any classes or functions you'd like to add to the (currently empty) beacon library.

That script should have the following characteristics:

1. Without additional arguments, it should retrieve the most recent event from the randomness beacon, and count the number of characters in the OutputValue the beacon returns. It should then print that output to standard out in comma-delimited format.

   For example, if the randomness beacon returned the string "01AF04F" for its OutputValue, your script should print:

    0,2
    1,1
    4,1
    A,1
    F,2
   

   The first value of any given line is a character. The second value is the number of times that value occurred in the OutputValue string.

2. When given optional --from and --to arguments, the script should output the character count over a span of beacon values using relative time. For example, if we provided the script the following arguments:

    $ summarize-beacon --from "3 months 1 day 1 hour ago" --to "1 month 1 hour ago"
   

   the script should parse the strings "3 months 1 day 1 hour ago" and "1 month 1 hour ago" into a time format the beacon's API can understand, and print the summed character count of all beacon values between those two times. The script should understand "month(s)," "day(s)", "hour(s)", and "minutes(s)", and should handle any combination of those values.

   Note that output is summed. For example, if the user enters --from and --to values that yield two beacon values, "B423AF" and "7352BA", the script should output the summed character counts of both values:

    2,2
    3,2
    4,1
    5,1
    7,1
    A,2
    B,2
    F,1
   

3. The script should error in a friendly way if a user enters a --from or --to time that predates the start of the beacon, Unix epoch time 1378395540.

A Quick Note: Don't feel obligated to solve this problem using just the python standard library. Part of engineering is knowing when to include 3rd party modules. In particular, if you know a good alternative to urllib or xml.etree, feel free to include it.

#Submitting Answers#

Submit solutions to this problem via Github pull requests.