Build your own simple Video Streaming service

Background: Four months back, when I was watching some video on Youtube, suddenly a thought came up to my mind about creating my pet project for video streaming service. I named it “Punahele” (Hawaiian word for Video :-)). What proceeded was an awesome development experience that involved using multiple technologies although the project remains incomplete.

In this post, I will walk you through the technical aspects of building a custom video service project like this one.

Overall setup & goals: Before we delve into the project, let me explain you the goals I had for this project.

·       The video should be streamed seamlessly to the user. At the same time, the CPU and RAM usage of the server should not shoot up high.

·       The users should be able to resume a previous session without any issue.

·       Video conversion process should not affect the users

·       A search interface should be provided to the user

·        Front End Module: This module is for the User Interface. The User Interface is mainly written in Bootstrap and JSP. I have chosen Bootstrap because it is easy to design box design. In Box design, the entire screen is divided into boxes. The outermost box is the entire page. This is divided into multiple sub boxes, one box contains Logo, Search box, Control boxes. Then another box contains the video thumbnails, video information and so on. The box structure is realized in Bootstrap using many inbuild css classes like “Container”, “col-sm-*” and so on.

·       
      Batch Conversion Module: This module runs as a Batch job converting videos into frames. We will see this in depth later.

·       
      Adapter Module: This module is packed with java interfaces for ElasticSearch, Redis and Kafka.

·       
      Core Module: This is the main module that will fetch required content and provide them to the front end.

Batch Conversion Process: The main idea behind converting video to frames is to have more control on the video portions requested by the user. This can also be capitalized to have better cache control on the video content. I have used JCodec library to convert video to frames. It provides a NIOUtils class that provides a readable channel for the file content. We then pass the channel to AWTFrameGrab class createAWTFrameGrab method that will return an iterator to the frames. We can convert the frames to a BufferedImage and write to a file using ImageIO class.

Here, once I am finished converting videos to frames, I write all the frames to single file and create a separate file that will hold the frame positions. I am doing this because this way, it becomes easier to read random file positions using RandomAccessFile.

Technologies: Core Java 8, JEE7, WebSockets, Futures, Kafka, Redis, ElasticSearch, Bootstrap, RandomAccessFile