- You can work in groups of 2 students.
- It is up to you if you want to fork this repo, or if you prefer to work in a private repo. However, you have to use exactly the same directory structure for the validation procedure to work.
- We expect that you will have more issues and questions than with other labs (because we have a left some questions open on purpose). Please ask your questions on Teams, so that everyone in the class can benefit from the discussion.
⚠️ You will have to send your GitHub URL, answer the questions and send the output log of thevalidate.shscript, which prove that your project is working in this Google Form.
This lab has 4 objectives:
-
The first objective is to design and implement a simple application protocol on top of UDP. It will be very similar to the protocol presented during the lecture (where thermometers were publishing temperature events in a multicast group and where a station was listening for these events).
-
The second objective is to get familiar with several tools from the JavaScript ecosystem. You will implement two simple Node.js applications. You will also have to search for and use a couple of npm modules (i.e. third-party libraries).
-
The third objective is to continue practicing with Docker. You will have to create 2 Docker images (they will be very similar to the images presented in class). You will then have to run multiple containers based on these images.
-
Last but not least, the fourth objective is to work with a bit less upfront guidance, as compared with previous labs. This time, we do not provide a complete webcast to get you started, because we want you to search for information (this is a very important skill that we will increasingly train). Don't worry, we have prepared a fairly detailed list of tasks that will put you on the right track. If you feel a bit overwhelmed at the beginning, make sure to read this document carefully and to find answers to the questions asked in the tables. You will see that the whole thing will become more and more approachable.
In this lab, you will write 2 small NodeJS applications and package them in Docker images:
-
the first app, Musician, simulates someone who plays an instrument in an orchestra. When the app is started, it is assigned an instrument (piano, flute, etc.). As long as it is running, every second it will emit a sound (well... simulate the emission of a sound: we are talking about a communication protocol). Of course, the sound depends on the instrument.
-
the second app, Auditor, simulates someone who listens to the orchestra. This application has two responsibilities. Firstly, it must listen to Musicians and keep track of active musicians. A musician is active if it has played a sound during the last 5 seconds. Secondly, it must make this information available to you. Concretely, this means that it should implement a very simple TCP-based protocol.
The following table gives you the mapping between instruments and sounds. Please use exactly the same string values in your code, so that validation procedures can work.
| Instrument | Sound |
|---|---|
piano |
ti-ta-ti |
trumpet |
pouet |
flute |
trulu |
violin |
gzi-gzi |
drum |
boum-boum |
- The auditor should include a TCP server and accept connection requests on port 2205.
- After accepting a connection request, the auditor must send a JSON payload containing the list of active musicians, with the following format (it can be a single line, without indentation):
[
{
"uuid" : "aa7d8cb3-a15f-4f06-a0eb-b8feb6244a60",
"instrument" : "piano",
"activeSince" : "2016-04-27T05:20:50.731Z"
},
{
"uuid" : "06dbcbeb-c4c8-49ed-ac2a-cd8716cbf2d3",
"instrument" : "flute",
"activeSince" : "2016-04-27T05:39:03.211Z"
}
]
You should be able to start an Auditor container with the following command:
$ docker run -d -p 2205:2205 res/auditor
You should be able to connect to your Auditor container over TCP and see that there is no active musician.
$ telnet IP_ADDRESS_THAT_DEPENDS_ON_YOUR_SETUP 2205
[]
You should then be able to start a first Musician container with the following command:
$ docker run -d res/musician piano
After this, you should be able to verify two points. Firstly, if you connect to the TCP interface of your Auditor container, you should see that there is now one active musician (you should receive a JSON array with a single element). Secondly, you should be able to use tcpdump to monitor the UDP datagrams generated by the Musician container.
You should then be able to kill the Musician container, wait 5 seconds and connect to the TCP interface of the Auditor container. You should see that there is now no active musician (empty array).
You should then be able to start several Musician containers with the following commands:
$ docker run -d res/musician piano
$ docker run -d res/musician flute
$ docker run -d res/musician flute
$ docker run -d res/musician drum
When you connect to the TCP interface of the Auditor, you should receive an array of musicians that corresponds to your commands. You should also use tcpdump to monitor the UDP trafic in your system.
Reminder: answer the following questions here.
| # | Topic |
|---|---|
| Question | How can we represent the system in an architecture diagram, which gives information both about the Docker containers, the communication protocols and the commands? |
 |
|
| Question | Who is going to send UDP datagrams and when? |
| Each musician will send a datagram every second | |
| Question | Who is going to listen for UDP datagrams and what should happen when a datagram is received? |
| The auditor listen to UDP datagrams. When a new instrument joins the orchestra, a new UUID is assigned to him by the auditor who will update his musician list. | |
| Question | What payload should we put in the UDP datagrams? |
| The only job a musician has to do is to play his instrument. So it will be a string of the sound of his instrument. | |
| Question | What data structures do we need in the UDP sender and receiver? When will we update these data structures? When will we query these data structures? |
| They will share the same datastructure : an array of objects. Each object is an instrument, having a name and a sound. They will query this datastructure to get the name from the sound and inversely. This array will not be changed. Moreover, the auditor will have an array of "active musicians", each active musician is an object having a "uuid", an instrument and a "first connexion time". It will be queried on each packet received to know if the musician is already known and active. The array will be updated on each first connexion : to add the musician and each time a musician become inactive (after 5s of inactivity) : to remove the musician. |
| # | Topic |
|---|---|
| Question | In a JavaScript program, if we have an object, how can we serialize it in JSON? |
We used JSON.stringify to convert it to a string |
|
| Question | What is npm? |
It is the package manager of Node.js. We used it to add new packages such as dgram to handle the UDP communication. |
|
| Question | What is the npm install command and what is the purpose of the --save flag? |
This option was used before npm 5to add dependancies, nowadays it is useless. |
|
| Question | How can we use the https://www.npmjs.com/ web site? |
| To search for the npm packages without having to install them directly. | |
| Question | In JavaScript, how can we generate a UUID compliant with RFC4122? |
By using the npm package : rfc4122 |
|
| Question | In Node.js, how can we execute a function on a periodic basis? |
By using the setInterval function with another callback function. But we have to manually call the callback function once before calling setInterval to not have to wait before the first call. |
|
| Question | In Node.js, how can we emit UDP datagrams? |
By using the package dgram, first we create a socket with : dgram.createSocket('udp4') and after we can send datagrams with the send() method on the created socket. |
|
| Question | In Node.js, how can we access the command line arguments? |
We used the process argument array : process.argv, it can be used like in C. |
| # | Topic |
|---|---|
| Question | How do we define and build our own Docker image? |
We defined the image in a Dockerfile and after that we build the image with docker build command |
|
| Question | How can we use the ENTRYPOINT statement in our Dockerfile? |
| It is used to run a command on a launched container. In our Dockerfiles we used it to directly run our node servers. | |
| Question | After building our Docker image, how do we use it to run containers? |
To run an auditor : docker run api/auditor. To run a musician : docker run api/musician <instument_name> |
|
| Question | How do we get the list of all running containers? |
By running docker ps |
|
| Question | How do we stop/kill one running container? |
docker container kill <container_name> |
|
| Question | How can we check that our running containers are effectively sending UDP datagrams? |
We can use tcpdump or wireshark |
| # | Topic |
|---|---|
| Question | With Node.js, how can we listen for UDP datagrams in a multicast group? |
When binding the connexion we can callback the function mySocket.addMembership(IP) |
|
| Question | How can we use the Map built-in object introduced in ECMAScript 6 to implement a dictionary? |
| We didn't used Maps but only standard js objects, so we can have the same datastructure on both side and not having to reversed maps. But if we used maps the sound search would be instant (with a hashmap). Because we only had 5 instruments we didn't needed the maps. | |
| Question | How can we use the Moment.js npm module to help us with date manipulations and formatting? |
We didn't used Moment.js but the standard js Date.toISOString() that was perfectly adapted. |
|
| Question | When and how do we get rid of inactive players? |
| When a musician become inactive, after not playing for 5 seconds. | |
| Question | How do I implement a simple TCP server in Node.js? |
We used the net package to create a server. The method on() listen to all connections and does all the necessary. |
| # | Topic |
|---|---|
| Question | How do we validate that the whole system works, once we have built our Docker image? |
We have used the validate.sh script, that runs multiple musicians and auditors. It checks if the musician are correctly added on first connexion (only) removed when they are inactive. |
Please be careful to adhere to the specifications in this document, and in particular
- the Docker image names
- the names of instruments and their sounds
- the TCP PORT number
Also, we have prepared two directories, where you should place your two Dockerfile with their dependent files.
Have a look at the validate.sh script located in the top-level directory. This script automates part of the validation process for your implementation (it will gradually be expanded with additional operations and assertions). As soon as you start creating your Docker images (i.e. creating your Dockerfiles), you should try to run it to see if your implementation is correct. When you submit your project in the Google Form, the script will be used for grading, together with other criteria.