Delivering Flawless Mobile Service at Large-Scale Events: A Cloudstreet Cookbook

By Mika Skarp

If you haven’t heard of it, ‘Slush’ is an annual start-up and technology focused event that happens in mid November in Helsinki Finland. Over two busy days, some 15,000 people congregate into one mammoth exhibition hall, providing fantastic showcase for technology and great place to network.

We had the great honor of being one of the companies on the floor, and took that opportunity to stage Cloudstreet ‘s official launch.  In short, it proved an amazing event in just about every respect. Every respect save one; internet bandwidth. The connectivity has always been bad, and this year it was virtually non existent.   So imagine yourself, the next great start-up entrepreneur, lucky enough to land some key VC meetings to demo your revolutionary App and the network is down!

Now imagine that the technology you are trying to demonstrate is actually THE solution for the “no internet” problem that’s keeping you from presenting. While it’s funny at first, the irony becomes, in a word, painful.

One of the cornerstones of the mobile network is ‘maximum capacity’. We use it to calculate “Erlangs” (unit of traffic density in a telecommunications system) for voice as well as data.  The standard approach to increasing capacity, and particularly in a situation like ours at “Slush”, is to add many small cells to cover the whole are. Nonetheless, if maximum capacity is not generous the network will quickly grind to a halt.

A second cornerstone is ensuring the network is clear of bottlenecks in backhaul or elsewhere in the network. In short, you should not make any statistical multiplexing for mass events, over dimension rest of the network compare to air interface.

A third cornerstone is the very orchestrated dance of managing users, and this is where 99% of networks fail (and Cloudstreet comes in). In a network with a rising number of simultaneous users allowed on the network, the capacity per user is diminished quickly,  can suffer dramatic variations and ultimately make the network unusable.  Much of the issue here is caused by retransmission of failed data packets, which comes to eat up end-to-end capacity. To solve this problem, you can limit number of simultaneous users in the cell, but that kind of first come first served network that would block new users from accessing it  is anathema to a public event like Slush.

To overcome this dilemma, and the challenge of managing multiple simultaneous users against maximum network capacity is to give users dynamically different user profiles.. With dynamic profiling, we can limit the maximum throughput of single user to ensure that the cell is not congested, while still providing sufficient levels of service. This means that retransmission of packages does not happen and the network will perform to continually fluid capacity.

Used profiles can be tailored per application over time. This means that when some applications needs more bandwidth for a certain time, this can be arranged by changing profile dynamically. At the same time, the network needs be “aware” of all other profiles in the network and ensure they are aligned with maximum cell capacity.  To do this, mobile network operators need to employ a kind of “token ring” type protocol in, where defined tokens are moved from one user to another based on need.