Deadline

6 Nov 2024

Introduction

• In this assignment, you need to write bit-rate adaptation algotithm.
• Skeleton code is provided and you need to make changes in only adapter.py. You can add your auxiliary files too.

Bit-rate Adaptation Algorithm

• We assume that video at the server is an array of chunks.
• These chunks are at different resolution in sorted order, therefore we can consider the video data as matrix.
• Algorithm takes input the state of client and manifest.
• Client call this algorithm each time it request next chunk. Given the current state of client, the algorithm returns the index of some resolution available in bit-rate resolution.
• Client then request the chunk with the resolution returned by Adaptation algorithm.
• The aim of this assignment is to maximize the reward/QoE(Quality of Experience) of the video.
• Reward Calculation : current_resolution - rebuffering penalty * (rebuffering) - smooth_penalty * (abs(current_resolution - last_resolution))

Simulating Video Playing

Client will calculate the delay to get a video chunk.
Drain delay from available buffer to simulate video playing.
If the delay is greater than avilable buffer then it will cause rebuffering, and rebuf will be max(0, delay - available buffer)

Case when requested chunk data + available buffer can overflow buffer threshold

Environments

We will use 4 different environments to test your algorithm. You can optimize for each one seperately. For each case, client will use seperate adapter written in respective classes.
Notation for classes :

• CC : Constant Chunk Length
• CB : Constant Bandwidth
• V : Variable

-- Channel Bandwidth in all cases will be 1000*1024 bytes/sec.

-- Server will have 8 Resolutions which are Multiples of 125 KBytes/sec.

-- Constant Chunk Length Size, Max Chunk Length, Penalties, Buffer Threshold are hidden for final leaderboard.

Constant Chunk Length, Constant Bandwidth

stateDiagram-v2
    Server --> Client : chunk fix length

Loading

Constant Chunk Length, Variable Bandwidth

To simulate the variable bandwidth, the topology will contain a randomized adversary which keep requesting packet of random resolution. It behaves similar to client.py except the adapter used will just return a random bitrate to be chosen.

stateDiagram-v2
    Server --> Client : chunk fix length
    Server --> Randomized_Adversary : chunk fix length

Loading

Variable Chunk Length, Constant Bandwidth

stateDiagram-v2
    Server --> Client : random_length_chunk

Loading

Variable Chunk Length, Variable Bandwidth

stateDiagram-v2
    Server -->Client : random_length_chunk
    Server --> Randomized_Adversary : random_length_chunk

Loading

Adapter.py

It contain 4 classes. You need to implement adapt method for each class. You are free to optimize for each conditions seperately.

What to Submit

• Don't change the name of files.
• You just need to submit adapter.py and auxiliary files on gradescope in a zip. Note that client.py and serve.py should not be submitted.