Pi Exchange

This project is software developed for the Raspberry Pi 3 model B to try and emulate a low-latency exchange simulation that can handle a large quantity of buy/sell transactions.

Getting Started

Both Python 3, C (has to be run on raspberry pi / Linux) and Java 10 compilers will be needed to run each individual part.

Linux:

Run the install.sh script to install the dependencies for the project, mainly a few libraries for python and Java 10 JDK.

Then run the run.sh script to run all the bits of the project together.

IF THE RUN SCRIPT DOES NOT WORK ON YOUR DEVICE DO: FOR THE SERVER

gcc tcp_server.c exchange.c order.c orderBook.c requestHandlr.c notification.c -o server -pthread

RUN WITH

./server

CONFIGURE THE GUI & BOT

When the server is running you will need the ip of the machine it is running on. You will need to introduce this ip in the “config.txt” file so that the GUI and Bot know where the server us running on.

RUN GUI WITH

python gui-v0.2.py

RUN THE BOT WITH Eclipse/IntelliJ or using other method to generate jar file

Easiest way is to use an IDE and run it straight from there, but you can also build the jar file if needed. Jar file should be kept in the same directory its source.

SERVER —> GUI —> Bot

Windows:

A powershell script is availible to install the python dependencies. Open powershell as admin and use Set-ExecutionPolicy Unrestricted to be able to run the script. Alternatively use pip install feedparser and pip install psutil. (Note that this will only work if you have a pip already or a version of Python that comes with it.)

The bot will need Java 10 JDK to be able to run on the machine you want. The server needs to be run on the Raspberry Pi using GCC.

Then just compile and run the bot, server and GUI. Take into account that you have to edit the port numbers and IP configurations manually.

Server

Compile: gcc tcp_server.c exchange.c order.c orderBook.c requestHandlr.c notification.c -o server -pthread

Run: ./server

Connect to server using ip and port 8890

Note: the IP depends on whether server is running locally or remotely. For local: 127.0.0.1, remote: find the IP of device where server runs on

Dependency: uthash.h https://github.com/troydhanson/uthash (src folder)

Place an order with 5 parameters divided by the single character “,”.

The sequence of parameters are shown below:

1. single character p (place)

2. b stands for buy, s stands for sell

3. price

4. amount

5. order id (generated by client)

e.g. p,b,1.1,10,0*

Cancel order with 4 paramter divided by as single character “,”. Sequence of parameters are as below:

1. single character c (cancel)

2. b stands for buy, s stands for sell

3. price

4. exchange id (generated by server)

e.g. c,b,1.1,1

Response from server

1. Status of current order

   • current order

   • order ID: %d

   • exchange ID: %d

   • price: %f

   • amount: %d

   • status: %d

2. Status of previous order

   • previous order

   • order ID : %d

   • exchange ID: %d

   • price : %f

   • amount: %d

   • status: %d

3. Best price

   • best buy : %f

   • best sell: %f

Note:

• to try the function of server, simple complie client.c with command “gcc client.c -o client” and run with command “./client 127.0.0.1” and type parameters as described above with enter
• current version of server only works well with 1 client, so please test with 1 client

Bot

Connect to server using ip and port 8890

Note: the IP depends on whether server is running locally or remotely. For local: 127.0.0.1, remote: find the IP of device where server runs on (ip address and port can be reset in config.txt under the same Java directory

Running the BOT

1. Initilization:

After running, this BOT will automatically start with generating initial buy and sell orders (total amount of buy and sell orders to be generated can be reset by modifying the value of Initialization.number_stored).

2. Providing market liquidity

The prices in this market is always changing when the BOT generate a new order. Initially:

* best buy price == Gprice-Math.random()-Ggap
* best sell price == Gprice+Math.random()+Ggap

By default, Gprice == 100 and Ggap == 1, which means the initial gap between the best sell price and best buy price is about 2. Due to the use of Math.random(), each order will likely to have a different price so that the price in the market is always changing. However, since the nature of trading is to “buy low, sell high”, the buy price will be always reducing, and the sell price will be always increasing. Hence, a gap between the best sell price and the best buy price is maintained: when the gap between these two prices > 3, this BOT will automatically cancel some orders and replace them with prices in the range of normal gaps.

How the BOT works

1. Sending and receiving orders:

All the orders must be synchronized and maintained in both the BOT and the server, which involves receiving / sending information of orders from / to the server. Two classes called Sender and Receiver are used to do the job. They are running concurrently by extending Thread, a mutex(lock) is used to allow them running in parallel.

Note:

• the string format of placing / cancelling orders is the same as mentioned above in the server part.
• since the information of orders is received in the data type of String, a class called Parse is used to extract and change each order’s id, price, amount, exchangeID etc. back to their original data type respectively (e.g. int) so that they can be stored in the BOT. The regular expression used is “\d+(\.\d+)?” to parse the String

2. Storing the information of orders:

A class called Store is used to save the information in each order, in the form of:

* id: %d
* price: %f
* amount: %d
* exchangeId: %d

Note:

• the constructor is: store(int id, float price, int amount)
• buy orders and sell orders are saved in 2 different Lists: buylist and selllist

GUI

Built With

Python 3, Java SE 9 and C compiled with GCC.

Authors

• Vasilis Ieropolous (Group Leader / Gui)

• Ke Chen (Administrator / Server)

• Francisco Caeiro ( GitMaster / Server)

• Julian Henjes (Gui)

• Guanghao Yang (Bot)

• Shen Huan (Bot)

  ---

  Thank you so much to Refinitiv for sponsoring the project.