You are here

T2.2 Probe Measurement and Analysis Primitives


Probes are deployed to provide a measurement platform on which measurement modules can be executed which ultimately will enable an analysis platform atop of these individual components. This task will take care of researching, specifying and implementing such modules. In particular, we will examine the following questions:

  • What kind of passive and active measurement primitives (algorithms) are useful for obtaining information about the network’s state when deployed in a distributed manner
  • What pre-processing of the above measurements is useful and feasible in a probe local context and how to dynamically trade-off pre-processing overhead with communication overhead.

A few examples of measurements that this task is likely to implement are:

  • Locally available configuration and operation data (bandwidth, link protocol information, L2-L5 error rates per service, etc.)
  • Path capacity estimation
  • Location
  • Round Trip Time and jitter measurements
  • Reachability and transfer tests
  • Path discovery, characteristics (e.g. traceroute)
  • QoE estimation (e.g., mimicking YouTube video player)

This task will also look into traffic classification techniques to identify frequently used services. Indeed, the knowledge of the applications generating traffic carried by the network is the first step to provide a fine-grained view of the status of the network. Given the current trend of moving services to the cloud and to use HTTP and HTTPS as protocol, and to the always-evolving set of applications, traffic classification is still an open issue. In the context of T2.2, both DPI and machine learning based traffic classification techniques will be defined to identify frequently used services.