Estimate the daily bandwidth consumed by YouTube globally.

Clarifications:

1. All formats and surface types / content types included = Yes

2. Youtube premium included = No

3. Includes watching and uploading = Maybe, lets see

4. YT music included = No

5. Global = Yes

6. Duration = Daily

facts about youtube (presumably known)

 

30M DAU (Google I/O announcement)

1 Billion hours of video watched per day (Google I/O announcement)

500 Hours video uploaded every minute à 500 x 60 x 24 = 720K = 0.72 Million hours of video upload per day (Google I/O announcement)

Ill answer this based on known stats about usage on YT and the split based on video type

Data usage by different resolution types

E.g., 480p video format has 30 fps and a bit-rate of 500 - 2000 Kbps this means on average 1250 Kb per second

1250 x 1000 bits / sec = 1.25 M bits x 3600 / 8 MB/Hr

= 0.5 GB / Hr

In similar fashion:

4K Video = 12GB/Hr

Full HD (1080p) = 2.5GB/Hr

720p (Standard) = 1.5GB/Hr

The over-all videos can be split as per table below:

Viewing and upload videos:

 

Resolution	% of content (guestimate based on what one observe)	Bandwidth / hour of video	Watch data (taking 1B hours of total videos/day)	Upload (ignore due to a factor of ~1500)
4K	1%	12 GB / Hr	1B * 0.01 * 12 = 120 PB	720 K * 0.01 * 12 = 86 TB
Full-HD (1080p)	19%	2.5 GB/Hr	1B * 0.19 * 2.5 = 500 PB	720 K * 0.19 * 2.5 = 340 TB
720p (Standard)	50%	1.5 GB / Hr	1B * 0.5 * 1.5 = 0.75 B GB = 750 M GB = 750 PB	720 K * 0.5 * 1.5 = 540 TB
480p (1250 Kbps, 30 FPS)	30%	0.5 GB / Hr	1B * 0.3 * 0.5 = 0.15 B GB = 150 M GB = 150 PB	720 K * 0.3 * 0.5 = 108 TB
Total			1520 PB / day	1072 TB = ~1 PB

 

Please note that total upload bandwidth is ~1500 times less than download and so one can ignore

Please also note that YT actually offers much more resolutions and fps (eg 360p to 8K), but this list is only an average and for a back-of-envelope this could be sufficient.

There are few additional factors we need to consider:

1. Buffering - Youtube doesnt directly 'stream' videos. It rather downlaods them in chunks. So, there is a possibility that youtube pre-downloads videos that are not "viewed fully" yet (typically a user discard a video in 10S but may be more than that is downloaded and buffered already)

2. Ads - There are many skippable and non-skippable ads that gets streamed before or within, that consumes bandwidth. Many of these ads are high-res and consumes more bandwidth than the video itself

3. Other in-app browing like thumbnails / previews. There are many who sees the previews before clicking the video. Its minor but at YT scale might be interesting

4. Assumption of 1Billinon hours per day of YT videos is pre-pandemic figure. There is all possibility that there is more surge on video watching now.

It might not be possible to attach a factor to each of these components, but I will assign a 30% buffer to count for these (specifically 1 and 4).

So total bandwidth / day = 1520 PB * 1.3 = ~ 2EB / Day which will be the ball-park for sizing

 

A thoughtful way to answer this Google estimation question -

Clarify : Talking about Youtube not Youtube TV etc. - Youtube site

Bandwidth Calculation: Avg Bandwidth used * avg # of hours of Videos watched per day per user * avg number of users per day

Certain assumptions - Check with interviewer

Bandwidth consumption may vary on certain factors, few important ones listed below and assumptions made

- Type of consumption device: Mobile, laptop, TV app, chromecast. Most Youtube videos consumed on phones/laptop. For the purpose of this excercise assume similar bandwidth on these devices  ( Assumption).

- Type of Video: Different type of videos can consume different bandwidth, Regular Video ( SD/HD), 360 Video, 3D, Live streaming. Most videos are standard video. Assumption  - Let us focus only on Regular Video.

- Type of usage could be Upload Video, watch Video, browse video, comment/like/share videos, subscribe to channels etc. Let us focus on the most bandwidth consuming and highest number of transactions i.e. watch/upload videos.

HD video 1 hr =  1GB; SD Video 1 hr = 500 MB

World Population : ~6B ( excluding China, Youtube access blocked)

Divided into 3 categories: Developed, Developing, Under Developed. Look at Internet penetration,  youtube usage, # of hr of usage, video bandwidth type. Video quality in developing and underdeveloped country might be lowere due to available internet speed.

Developed Countries ( 20% Population) = 1.2B --> Internet Penetration (90%) --> 1B, Youtube Usage --> 60% ppl --> 600M

Avg Use per person per day: 2hrs of HD Video i.e. 2GB; Total Bandwidht used per day = 600M * 2GB = 1200 Petabyte

Developing Countries ( 60% Population) = 3.6B--> Internet Penetration (60%) --> 2.1B, Youtube Usage --> 40% ppl --> 800M

Avg Use per person per day: 1hrs of SD Video i.e. 0.5GB; Total Bandwidht used per day = 800M * 0.5GB = 400 Petabyte

Under Developed Countries (20% Population) = 1.2B--> Internet Penetration (30%) --> 360M, Youtube Usage --> 30% ppl --> 100M

Avg Use per person per day: 30hrs of SD Video i.e. 0.25GB; Total Bandwidht used per day = 100M * 0.25GB = 25 Petabyte

Total bandwidth Usage globally per day = 1200+400+25 = 1625 Petabyte

Clarifying questions:

1. For US or for the global user base? - Global

2. Is this for Youtube on mobile and WWW? - Yes both

3. Does it include YT red - no

4. Per day, per month or annually. - Per day

Equation = # of videos seen per day * avg size of each video -- (1)

Assuming that all the users watch a video of 10 mins length on an average.

Split 1:

(# of videos seen per day in HD  * avg HD size of each video) + (# of videos seen per day in SD * avg SD size of each video)

Average video size of HD for 1 hour - 1 GB, 10 mins → 100MB

Average video size of HD for 1 hour - 500 MB, 10 mins → 50MB

Estimate the number of videos seen across the world

World Population = 7B. 50% internet penetration → 3.5B users

20% of world lives in developed world → 700M

60% of world lives in developing world → 2B

20% of world lives in under developed world → 700M

Number of users who visit YT in developed countries → 60% visit per day → 420M

Assuming on average a user watches 5 vidoes on average per day → 5 *420M videos seen per day → 2.1B

Of these, it is safe to assume 20% watch HD videos and rest in SD → 420M videos

80% watch SD → 820M videos

 

Number of users who visit YT in developing countries → 50% visit per day → 1.1B

Assuming on average a user in the developing country watches 3 videos on average per day → 3 *1.1B videos seen per day → 3.3B

 

In a developing country, it is safe to assume 10% watch in HD and rest in SD

HD video seen per day → 330M

SD video seen per day → 2970M

 

Number of users who visit YT in underdeveloped countries → 20% visit per day → 140M

Assuming on average a user in the developing country watches 1 video on average per day → 1 *140B videos seen per day → 140M

In a developing country, it is safe to assume 10% watch in HD and rest in SD

HD video seen per day → 14M

SD video seen per day → 126M

Total HD videos seen:

420M+330M +14M ~ 800M

Total SD videos seen:

820M+2970M+126 ~ 3800M

Total Bandwidth Consumed thus (assuming average video length is 10 mins):

(800M*100MB) + (3800M*50MB)

(8*10^16) + (19*10^16)

27*10^16Byter

27Petabytes