Why is my throughput so low?

You are responsible for deploying a wireless broadband access network. The network is heavily used and is therefore being upgraded. You just received a shipment of YMax-II basestations to replace your old YMax-I basestations. The new model is very similar to the old one but it has a 70 MBPS max data rate, compared to 17.5 MBPS of the old model. You are looking forward to approximately quadrupling the throughput. Currently the average throughput of each basestation is only little over 10 MBPS. The new basestations go on-line and everything is fine, except that the average throughput only went up to 12 MBPS – a far cry from a factor of four! So what went wrong?

Unfortunately nothing went wrong. You just ran into the difference between throughput and maximum data rate. Here is a brief explanation of what is going on.

The first thing you need to understand is that the data rate which can be sustained depends on the signal to noise ratio (SNR) at the receiver which in turn depends on the distance. The further away you are from the basestation, the lower the SNR and the lower the data rate. The following figure illustrates the dependence of rate (the vertical axis) on range (the horizontal axis) for the YMax-II system.

rate_range.jpg

Wireless systems such as YMax (as well as WiMax, WiFi, 1xEV-DO, WCDMA etc.) are able to adjust their rates to accommodate users located in different places in the cell. A user at a distance of 200 meters from the basestation will get 70 MBPS while a user at 600 meters will get approximately 7 MBPS. If we design our deployment to have small cells of 200 meter radius or less, all users in the cell can be served at 70 MBPS. However, we will need a lot of cells which is prohibitively expensive. Instead we designed our system to have larger cells with 600 meter radius, which means we need approximately 9 times fewer cells, greatly reducing the cost of the system. The consequence of this is that users at different locations in the cell get different data rates.

What we are interested in as the owners of the network is the average data rate or throughput we can “push” through a basestation. Assuming that users are randomly distributed in the cell and that each user is promised equal service (say, an average data rate of 250 KBPS), we can compute this throughput. I will omit the details of this calculation and just present the result in the following figure. Here the vertical axis is the throughput and the horizontal axis is the radius of the cell. Because our cells have a radius of 600 meters this figure tells us that the throughput is 12 MBPS which matches what we observed. A similar calculation for the YMax-I system show that the throughput is approximately 10 MBPS.
throughput.jpg

Why is the throughput so low? Users in the cells are served at data rates between 70 MBPS (near the basestation) and 7 MBPS (at the edge of the cell). The throughput is much closer to the worst rate than to the best rate. There are two reasons for this fact. The first is that low data rate users take up the bulk of the resources of the basestation. The basestation needs to transmit to a 7 MBPS users for 10 times as long as to a 70 MBPS user to deliver the same average number of bits per second. Thus the basestation spends most of its time handling the users with the lowest data rates. The second reason is that there are many more low data rate users in a cell than high data rate users. This is simply a matter of geometry – the area near the cell edge is larger than the area near the basestation. The consequence of this is what we see in the figure.

The moral of the story is that the peak data rate of a wireless system tells only part of the story. Unless you look at the complete picture you may reach very wrong conclusions about the impact of increasing the peak data rate on the the actual amount of data which can be pushed through a basestation. In this example you expected to quadruple the throughput and instead got less than 20% improvement. Well, that’s reality for you.

Of course when the vendor came to sell you the upgrade all he talked about was how wonderful it is that the new system is four times as fast as the old …

Advertisements

2 responses to “Why is my throughput so low?

  1. Is your speed-vs-distance curve based on interference or just thermal noise? In CDMA, which is usually interference limited, users close to a cellsite get much better service mainly because interference from neighboring cells is so much less. It also means that while subdividing cells will increase total system capacity, it won’t necessarily increase the average per-user speed.

    Cell subdivision also hits diminishing returns as they get close enough to be in line of sight to each other. Then the propagation becomes inverse-square rather than inverse r^4 and Olber’s Paradox becomes a problem.

    I think it’ll take more than just lots of cells to provide very fast wireless data services; at some point, better and more clever use of antennas will become key.

    Or we could just pull fiber to every home and be done with it. Guided waves beat unguided waves every time. 🙂

  2. JimDeGries@gmail.com

    Phil

    The figures were computed assuming thermal noise only, ignoring inter-cell interference. Think of YMax as a system using a high-factor frequency re-use. In CDMA and some other systems we have frequency re-use of 1, and the problem is interference dominated. In this post I tried to explain the simpler version of the problem, and make the point that throughput is very different from peak data rate. This is true also in the interference dominated case, although the numbers come out quite different. To explain something relatively complex in a brief post one must simplify.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s