4.7 KiB
Links
Project Description
This project utilizes the steam API and graph databases to create friend graphs for clients in a web browser. Currently there are two types of graphs available:
- Friends of Friends Graph: This graph displays all the steam friends of a single person and all their friend’s friends.
- Common Friends Graph: This graph will only display your friends; however, it will draw edges between your friends’ if they are friends with each other.
Motivation
While learning about graph databases I thought that it would be awesome to create a massive graph database with steam friends in it. After a quick google search, I realized that some other people have done similar things in python, however, nobody has ever made a live website do this. Using my prior knowledge of the steam API and new knowledge of gremlin, Hadoop and HBase, I figured that this would be a great weekend project. Although I got most of the functionality done in the first weekend, I ended up working on it for 3 weekends hammering out all the edges. After ample work, I am satisfied with the appearance and functionality of the website, but, I am looking forward to making some improvements in the future.
Technical Details
At the core of this project there is a java backend and a JavaScript frontend. I needed a backend for this project because you cannot distribute your steam API key, plus, this allows me to cache everything in a local graph database. The second time a player is queried, it is extremely fast since all the steam friends are cached in a local gremlin server. The only time-consuming operation that the backend must carry out is querying the steam API; querying the gremlin server is relatively fast. On the front end I used Sigma-JS with a web socket to render the steam graph. Since creating a new graph on the back end can take up to 5 minutes (if the player has not been indexed), I continuously send new nodes to the client as they arrive from the steam API. This prevents the client from staring at a loading bar for 5 minutes; plus, it makes great time lapse videos. The only drawback to this is that you cannot start applying a force to shake the graph until all the nodes are added to the graph.
Uses of Friends Graph Data Base
There are tons of academic papers written on the usages of friends’ graphs. What you may be able to do with friends graph may shock you. A popular use of friends’ graphs is to view friends’ clusters. Facebook may be able to identify someone's family, friends, and work groups by simply using their friend’s information. When looking at steam friends graphs I could identify some people's high school and college friend’s groups. Another famous example would be for friends’ identification; If a ton of people in your friends list are also friends with a person, it is likely that they would also know that person. This is how websites like Snapchat, Instagram, and Facebook recommend people for you to follow. Similarly, you can use this to identify people’s best friends. If two people share a ton of friends, it is more likely that they will be good friends vs the person which they only share one friend with.
Simple analysis of graphs like this may seem benign. However, there are tremendous implications when it comes to privacy and advertising. If websites chose to, they can harvest a ton of data and start recommending you products based on not just what you like, but what your friends also like. Steam already does this to a certain extent when they recommend you games that are popular among friends.
Hosting the Project
Hosting the front end of this project is easy, it is just thrown on an apache server. The backend is trickier since it needs to run both a java app and a gremlin server. Currently I am hosting this project on an Ubuntu VM with 4gb ram. At idle this project only takes 1gb of ram, however, under heavy loads it uses around 2 GB of ram and saturates my CPU.
Future Plans for the Project
In the future I would love to add more graphs to this website. Instead of just displaying graphs it would also be cool to create pages which tries to identify friends’ groups, people you may know, and best friends. There is a ton of game information available on the steam API, it might be neat to make graphs based on friends who play the same games.