Table of content:

• Functional Requirements
• Non Functional Requirements
• Capacity Estimation
• High Level System Design
• Algorithms

YouTube or Netflix are popular video streaming services. These services allow users to upload / stream /share videos.

Functional Requirements

1. User should be able to upload and stream videos
2. System should support different video formats
3. System should recommend videos to viewers

Non-Functional Requirements

1. System should highly reliable/scalable
2. System should be highly available with eventual consistency

Capacity Estimation

Throughput

• DAU : 5M users and each user watch 5 video/day
• 25Million views/day-> 250 Views/sec(RPS)
• 10% of users upload-> 5 videos/sec

Storage

• 5M X 500MB= 250 TB/day

High Level System Design

• User uploads videos into object storage and updates medata into database(Cassandra)
• Transcoder service will read from queue process them asynchronously and encodes into different format.
• Different video formats make it available to CDN to improve view experience.
• Video recommendation service uses two-tower approach to generation video recommendations.
• Client stream videos on their devices using different streaming protocols such as Apple HTTP Live/MPEG.

Algorithms

Below algorithms helps to find top K movie recommendation using Heap DataStructure.

from heapq import *

class ListNode:
def __init__(self, val=0, nextNode=None):
self.val= val
self.next = nextNode


def topKMovies(arrays):
heap = []
for head in arrays:
heappush(heap, (head.val, head))

    topMovies = []
    while heap:
        nodeVal, node = heappop(heap)
        topMovies.append(nodeVal)
        if node.next:
            heappush(heap, (node.next.val, node.next))
    
    return topMovies

if __name__=="__main__":

    n1 = ListNode(1, ListNode(5, ListNode(7, None)))
    n2 = ListNode(2, ListNode(4, ListNode(8, None)))
    array = [n1, n2]
    print(topKMovies(array))