Can You Queue Video Uploads to Youtube

YouTube Upload Processing has many intriguing queueing properties.  This article volition primarily expect at the mpeg-to-swf conversion and written report out the queueing properties of that process. Y'all can likewise view all forty+ articles on Queueing Theory.

>

I'll prove the bones process of how to upload a video on Youtube, explain the Queueing mechanics that goes on backside-the-scenes, propose a few books in case you're interested in Queueing Theory applications, and and so finally show a video tutorial on an introduction to Queueing Theory and Models.

I make many assumptions in this article, so feel free to poke holes in what I'chiliad arguing.

Assumptions and Data

Below is a very elementary, loftier-level process map of the steps to upload a video on YouTube:

In summary, the user does the following:

1. User uploads video
2. User waits while video is uploaded
3. Upload is completed
4. User waits for video to exist converted from MPEG to SWF
5. Conversion from MPEG to SWF completes

Based on a small sample of 20 uploaded videos of an average file size of iii.5 MB, I summate the mean for uploads to be ~180 seconds per MB and ~240 seconds per MB for the conversion.

Based on YouTube's own disclosure, nosotros also know that there are on average 65,000 video uploads per day.  Nosotros do not know the average file size.

Queueing Properties

Important items to note when studying the queueing backdrop of a system are the following:

1. λ = Arrival Rate, or more specific, the fourth dimension between arrivals.  For nearly queues, we can assume that the inflow distribution tin can exist approximated by a Poisson distribution; which means that the time between arrivals are non deterministic, but random.
2. μ = Service Charge per unit, or more than specific the time for a arrival to be serviced.

A poisson distribution typically looks skewed to the left or to the correct — that is because the mean and the standard departure is the same.  Here's a standard movie of a poisson for server utilization:

What we see above is that as there are more simultaneous connections, at that place is a subsequent arrival rate batching — represented past the poisson curve above.

Given the information and notation, we can now attempt to better understand the queueing properties of the MPEG-to-SWF conversion. Remember: the data I have assumes several things and, is virtually likely, completely off the marking.  Only, it'southward an endeavour and, if anything, it's fun to endeavour.

Presuppostions Redux

So, 65,000 video uploads on a 12 hour day, gives us the following:

λ = 65,000 / 720 minutes = (90 / minute)
μ = 3.5 * 240 seconds = 840 seconds; 840 seconds / 60 = (14 conversion / minute)

Arguably, fourteen conversion / minute is very depression.  Let'due south just assume that YouTube average service charge per unit is 200 conversions / infinitesimal.   Given that, nosotros can now larn nigh the queueing backdrop, which I describe below.

Boilerplate Number of Videos Waiting to be Converted

The equation to learn about the average number of files in the conversion process is the following:

C_westward = (λ^two / μ(μ – λ))

And then,

C_{west =} [(8100) / (200(200 – 90)] = .36

So, given the assumptions above, non fifty-fifty 1 video is waiting to exist converted from MPEG-to-SWF.

Boilerplate Number of Videos in the Conversion Process

C_s = (μ – λ)

So,

C_{s =} [(xc) / (200 – 90)] = .81

This means, given the assumptions above, that an whatever betoken in time, there is ane video in the conversion system.

Average Time Spent Waiting

T_west = (λ / μ(μ – λ))

And so, nosotros become:

T_{due west =} [(90) / (200(200 – xc))] = .004

This ways as videos enter the conversion process, there is inappreciably whatever waiting — they are served almost immediately.

Boilerplate Time Spent in the System

T_s = (1 / (μ – λ))

So, we get,

T_{s =} (1 / (200 – xc)) = .009

This ways, and so, that as videos are uploaded and enter the conversion queue, they are served nearly immediately, without any waiting.

Weaknesses in Assay

Okay, the numbers above are pretty much pulled out of the clouds.  Merely, if we had real information, then you could just plug them into the equations above.  My gauge is, though, that even if we had existent numbers, the results would be really close to what I show above.  Why?  Well, for each input into the YouTube system, ane could argue that information technology has very little affect on resources — this is a common property in telephony and in server modeling.  I encounter the same thing going on with YouTube.  The biggest claiming for YouTube is not computing resource, but storage capacity.

larimoretoppland.blogspot.com

Source: https://www.shmula.com/youtubes-queueing-properties/349/

Share this post