Some of you may have heard about Pokemon Go – it is already being proclaimed the most successful mobile application of all time (at least from a launch week perspective). A few of the stats that I have seen are amazing (and there are many, many more):
- Over 33 minutes per day – more than Facebook
- 21M daily active users
- Two days after launch, already on 5% of Android phones in the US
We have all seen people walking around like zombies playing the game. But how much impact is this having on the mobile networks that you are on?
I decided to take a look at one mobile network and see what impact it was having, not just on bandwidth, but also on the subscriber participation percentage as well as the number of sessions that the application generates. And as I suspected, it was very interesting….
I looked at a European mobile network that was almost 2M subscribers , and over a three hour period, over 7% of the active subscribers on the network played Pokemon Go.
That is pretty amazing for such a new application, and shows how quickly that a new application can affect networks. The above chart demonstrates that there is not a lot of volume per subscriber for the application, and at least on this network, the quality and latency are pretty good. Looking at the traffic however shows that the application really does not generate a lot of traffic per subscriber (as shown in the graph below), so that means that the application doesn’t really hurt the mobile network, right?
Not so fast….
Bandwidth is just one factor on a network. One usage parameter often overlooked is the number of sessions that the application generates, which affects “stateful” network elements (analytics, charging, and security among them), have to deal with “chatty” or “noisy” applications as they consume system resources.
Pokemon Go is already pretty chatty – the real-time screenshot below shows that although the bandwidth is pretty low for the application (about .1% of the overall traffic) compared to Facebook (which is about 16% of the overall traffic), the session total for Pokemon Go is almost 1% of all sessions on the network.
That may seem harmless, but remember that Pokemon Go has not yet fully monetized their model with advertising and sponsorships. So take that number, and then imagine that every 10 feet someone walks, a new sponsored ad pops up, or a company is paying Niantic to let them know when Pokemon players are near their location, generating a new session and data that gets sent back to the game servers. This may would likely pop the total up nearly exponentially, which would mean that mobile operators would have to do capacity planning with this as a major factor.
The chart below compares the number of connections per subscriber that Pokemon Go is generating compared to other applications. As you see below the only application that generates more connections per subscriber is BitTorrent file sharing – which is notorious for session consumption. Pokemon Go even uses more connections per subscriber than DNS and SSL web browsing, which are the meat and potatoes for any device connection to the Internet.
This was just a quick review on how Pokemon Go is hitting broadband networks – We will keep an eye on how the app evolves as it starts adding more advertising to see if that dramatically changes the behavior.
This is a good example of how application-aware analytics can help an operator react to something unexpected like Pokemon Go. Interested?