Real-Time Prediction of Traffic Statistics and Patterns in Programmable Networks using Probabilistic Data Structures


Scope: Communication Networks
Main KTH Staff Involved: Prof. Rolf Stadler
Other Partners: SAAB (Stefan Hagdahl and Mats Jonsson)
Project Idea: To support network applications and their distinct requirements, such as low latency, high
throughput, low jitter and small error rates (high reliability), the network infrastructure needs
to be adaptative, programmable and elastic. It is important to build intelligent mechanisms that
allow the automation of network management tasks aiming to simultaneously obtain the
desired performance.

The network management task is enabled by the analysis of good performance indicators,
which allows network administrators to evaluate the quality of the services provided by the
network. In this context, network monitoring is a key aspect of network management. In
general, it can be said that network monitoring is the process of taking measurements so that
an abnormal behavior can be inferred. Knowing this information, it is possible to carry out
capacity planning, fault diagnosis, anomaly and attack detection, and execute an optimized
traffic engineering to improve applications’ experience in the network.