• Twitter
  • Facebook
  • Google+
  • LinkedIn

FairNet: Measurement Framework Traffic Differentiation Detection on the Internet

FairNet: Measurement Framework Traffic Differentiation Detection on the Internet

As the Internet evolves as a preferred platform for many commercial activities, the guarantee of high-performance is expected from Internet service providers (ISPs). Often ISPs themselves are content providers (CPs), and would like to ensure higher performance guarantee to their content than that of their competitors, or, in the worst case, can even degrade competitors performance to capture higher market share. There have been cases of ISPs giving preferential treatment to specific CPs, which has prompted regulators worldwide to advocate Network neutrality. To enforce any neutrality laws, one has to have a mechanism to identify any violations. Our work proposes a method to detect any deliberate discrimination of content of a particular service on the Internet.

Network neutrality is related to `equal' treatment of the same type of traffic (video, audio, file transfer) on the Internet irrespective of their origin and destination.  Most regulators allow some traffic discrimination under the 'reasonable' traffic management. Note that there is no difference between such traffic management and traffic differentiation from the implementation point of view. Moreover, the multiplicative increase in the window size of the Transmission Control Protocol (TCP) often oversubscribes the Internet link leading to congestion. The difference in the geo-location of servers, providing content to users, imparts unavoidable variations in the congestion levels. From the user point of view, the effect of congestion is also no different from traffic differentiation.

We mainly focus on the traffic differentiation of multimedia streaming traffic, like YouTube, Hotstar, Netflix, Amazon prime video, Gaana, Saavn, Spotify, etc. However,  our method applies to any type of traffic. The multimedia streaming traffic is characterized by varying data rates controlled by Dynamic Adaptive  Streaming over HTTP (DASH) mechanisms. DASH could be proprietary for services. The effect of rate control by DASH is also perceived as performance degradation on the user side. If one notices any performance degradation of a specific traffic stream (say YouTube), it is hard to pinpoint its cause. Instead, the performance can be compared to single out deliberate performance degradation.

Our measurement framework, named FairNet, ensures that the aspects mentioned above affect all the traffic streams, in the same manner, imparting the same level of performance degradation. If there is deliberate degradation of the quality of a traffic stream, only that stream's quality further degrades and could be identified by performance comparison. FairNet framework consists of a user client interacting with the common streaming server (CSS) that hosts the data associated with all supported services. The common streaming server primarily removes the variations in service throughput performance due to congestion and dynamic adaptive streaming techniques. The service type selection on the user side manages the disturbances due to network traffic management techniques. The 20Mb of original streaming service data is used by the CSS. This data is divided into smaller data chunks called segments. The server runs the customized, dynamic adaptive streaming mechanism that performs rate adaptation for each data segment as per the perceived channel conditions on the server-side. We developed an Android App named after measurement framework as "FairNet App" for users. 

It is available on the Google Playstore. The correct setting of protocol parameters and its sequence ensures that the traffic generated by FairNet is classified as respective streaming service traffic while traversing over the Internet. This aspect is validated using commercial traffic shaper in the lab.

We developed a novel traffic differentiation detection algorithm that uses a time-windowed throughput comparison, socket connection status, and downloaded data size. The lab experiments, as well as field testing, has established the correctness of these algorithms. 

Thus, our FairNet framework can generate comparable throughput performances and identify deliberate traffic differentiation on the Internet.  The user can check if any of their favorite services is deliberated throttled in the network by using the FairNet App. Hence it acts as a deterrent for any net neutrality violation on the Internet.