How to calculate your bandwidth use for streaming media hosting

How to Start?
Most people who are evaluating the use of the GravityLab’s Media hosting for audio and video will reach a point where they have questions along the lines of the following:
“Which streaming media hosting plan should I choose?”, “should I use Windows Media, Quicktime, Flash or some combination?”, “how much storage allocation will I need?”
Calculating Simultaneous Users:

Most companies do not have accurate estimate of the number of simultaneous users on a particular portion of their web site. In general, web stat logging software typically does not provide this much granularity. In fact, if your client tells you that they expect a hundred thousand, a half of million, or even a million or more simultaneous users, it is probably time to raise a red flag.

Unless you are building the next version of Yahoo’s IM, MSN’s IM, or AOL’s IM, your web site will not see this amount of simultaneous traffic. Chances are your client was talking about the amount of total page views or hits over the course of some time frame.

For this reason, it is important to understand what we are looking for. The number of simultaneous users that will be using your application at peak load is the maximum amount of users that will be using your application at the same time. It is not the total number of users who will ever use your application. It is not the number of people who have used your application in a single day. The number of simultaneous users that will be using your application at peak load will likely be much lower than the total number of users of your application.

How to Estimate Number of Simultaneous Users:

Since this the number of simultaneous users is such an important piece of information for determining which hosting plan you choose for your account and the bandwidth capacity you will need for your application, we will walk you through one way of estimating this number when the only information you have is very general web page statistics.

We’ll use a hypothetical case: The ACME Company intends to launch a 10 minute marketing video linked off the home page of their web site for 20 days to promote an upcoming product launch. They have stats from their last product launch that showed 150,000 people viewed the promotional materials during that 20 day campaign. In this example, we also know that almost 50% of this traffic occurred in the first 5 days of the campaign.

To find an estimate for the number of peak simultaneous users, we’ll try to find the single hour in which the most users viewed the video and then make an assumption that 1/6 of the visitors in that hour will be watching the video simultaneously. The 1/6 factor is a simplifying assumption based on the fact that the viewers will be distributed evenly in that peak hour and each will be viewing the video for only 10 minutes of a possible 60 minutes in the hour. Therefore at any one point in the hour, on average, only 1/6 of all of the viewers in that hour would have been viewing the video simultaneously (given a random distribution).

In the example above, since we don’t know how many visitors our site will have during the peak hour, we’ll need to make some more assumptions. First, since we know that about 50% of the traffic will occur during the first five days, let’s assume that there is a spike in traffic during the first five days, and then traffic gradually decreases. With these assumptions, and the fact that we expect 150,000 viewers over 20 days with about 50% of those in the first five days, we can use Microsoft Excel to model up an estimate of daily web traffic.

In this example, our model would suggest that the day with the largest amount of viewers would be Day 4 with 16,250 viewers. The next step is to find the hour with the greatest amount of expected viewers. Making the assumption that this marketing campaign will be aimed predominately at US audiences, we can use Excel to develop a similar model of estimated traffic per hour during the day.

The distribution of traffic into the various hours as shown above roughly models the traffic we’ve seen for certain videos deployed on macromedia.com. Using this model with our ACME Company example we’ve now identified that the single hour during the entire campaign (20 days) with the most viewers will be from 11AM – 12noon PST on Day 4 and that there will be 1,950 viewers at this time.

We can now multiply the 1,950 viewers during the peak hour by 1/6 to get the estimate for the peak number of simultaneous users. In this case, our estimated number of peak simultaneous viewers of this video will be 325.

Examples of Calculating Simultaneous Users

In this example, the application is a one­way data­casting system that streams real time monitoring info to a group of end users. In this example, the Media Server will serve only one stream to each user.

Simple Connections Formula

A simple way to calculate simultaneous users can be made with the
following data points:

Total Users over a known time period

Average Connection Time for a user (normalized)

Formula:

U(s) = U(T) / (T / A)

U(s) = Simultaneous users
U(T) = Total users
T = Time period
A = Avg. connection time

0