Soda-Pop-Refresh is a webserver stack running on a Raspberry Pi 3 that pulls data from a vending machine on Calvin College's campus (SB 382) and generates a webpage for the use of students and faculty.
The vending machine, a Vendo 540 model, has a simple microcontroller that keeps track of various data-points such as sales, temperature, money in the machine, and other diagnostic data. This device can then offload the data over a serial interface using the DEX protocol. This data is normally managed by specially designed devices used by a vending technician whenever they refill or service the machine.
By creating a 1/4" TRS to RS-232 (9-Pin D-Sub) cable, we can pass the DEX data into our Raspberry Pi. This is handled by the C code in dex_communicator; which drops the DEX data into a temporary file. Our glue code, written in Python, then passes this DEX data to dex_parser, which converts the DEX to XML. This XML is then parsed and added to an sqlite database.
We chose to host the web server on the Pi itself. Because of this, the server had to be small in size and fast. As we want to build a webpage populated with information from our sqlite DB, we also wanted to use CGI (Common Gateway Interface) to allow our scripts dynamic generation ability.
Given these constraints, we chose Lighttpd as it is light and small (unlike apache), and it supports CGI out-of-the box (unlike nginx).
Our lighttpd server is barebones. Currently we configured it just enough to pass through to our python scripts that we put in /var/www/cgi-bin/.
The progress we made from a very outdated server running on a different architecture and a full dedicated machine, to an up-to-date stack acheiving the same responsiveness while being contained entirely on a credit-card sized ARM SoC, is a great start. However, this project is not done as of today.
Calvin's C.S. Club, Abstraction, has adopted this project and intends to add the following by the end of the semester:
- Responsive Website Design
- Social Media Integration (Tweet whenever it sells a soda, for example)
- Internal Management Tools and Statistics
- To Elaborate on this a little, we would like to use the sales data with a statistical model to give the soda-pop refillers (student volunteers!) an estimate of how much of each beverage we should get. Once we have 365 days of data, we could even pre-emptively purchase more stock during seasonal highs (e.g. finals) and reduce refill frequency during lows (e.g. summer).