Control Traffic Overload in Mobile Networks
ParidhiVerma 2700008QKP Visits (5089)
It is an open secret that the mobile applications and mobile devices are growing in an explosive manner. And it is also well known that the growth of mobile data is causing a tremendous problem within the mobile operator networks. As an example, Cisco released a study which showed that mobile data is tripling every year.
The growth of mobile data brings to the forefront one of the basic design choices that network protocol designers had to make during the protocol development, specification and standardization. That basic design choice is one between that in-band control and out-of-band control. But what really do these two design choices mean?
The primary objective of any computer network is to allow devices to talk to each other. The information they exchange with each other, e.g. getting the bytes that let you see a movie or hear a song, is called the data traffic. But in order to get the data traffic to flow, the network needs to do some housekeeping chores underneath, e.g. let one device find where the other device is, set up communications channels between them and other things of that nature. That traffic is referred to as control traffic.
One set of network designers believed that control traffic was a totally different beast than data traffic, and that the two should be separated from each other in the network. The logic was that control traffic would be isolated and protected from the massive fluctuations and vagaries of the data traffic. The traditional telephony network and most of the mobile network protocols were designed with this philosophy, which is out-of-band control. Another set of network designers believed that there was no reason to separate those two types of traffic. That is in-band control, and you can see this design philosophy within the TCP/IP protocol suite that emerged out of the IETF.
Mobile network protocols have used out-of-band control, and mobile network operators provision separate channels to carry control and data traffic on wireless links. However, applications on smart phones run on top of the TCP/IP protocol suite, and programmers of such applications are oblivious to the fact whether their data is causing a control traffic flow or a data traffic flow on the wireless network. These idiosyncrasies of the underlying network are hidden from the programmers by multiple layers of software and protocol specifications. As a result, mobile network operators are discovering to their immense displeasure that smart phones can generate up to ten times more control traffic than regular phones for the same amount of data traffic. What this means is that the control channels on the mobile network is likely to get overloaded faster than their data channel get overloaded as more and more smart phones come online.
So what is the solution out of this sticky situation? The operators can provision more resources for their control traffic, but they don’t particularly like that solution. There is some justification for that since users pay for data traffic, and not for control traffic. The operators also can’t migrate over to an in-band control model since that design philosophy is baked into the network protocols standards.
That leaves only two possible solutions on the table, (a) have the smart-phone applications be written so that they do not cause too much control traffic overload into the network or (b) deploy some appliances that are located within an operator network and protect the older equipment in the mobile network from the onslaught of the control traffic. In order to attain (a), a set of best practices and guidelines to minimize control traffic load on the network can be incorporated into the development process of mobile applications using software engineering tools. Different types of appliances can be imagined for (b) located at various points in the mobile network. Some progress towards (b) is already being made by startups such as Genband which provide an ability to mitigate the signaling overhead due to the smart phones.
In the longer term, one does hope that the future generations of mobile network protocols designers will develop new protocols that use in-band signaling. The success of the Internet protocols and their triumph over various telephony protocols shows that this approach might indeed be the better one to use for improved scalability.
About the author
Dinesh Verma is a Researcher and Department Group Manager at