The submission deadline:

Join us Vkontakte and follow news in Russian

Welcome to visit Open Karelia

You are here

Load balancing in mesh-networks based on process mining algorithms


Background and Motivation: 

In spite of an essential progress in wireless networks standardization, a set of unresolved technical tasks still exists. One of such a task is traffic forecasting and management. Historically, routing protocols and algorithms came from wired networks where network topology and data flow  structure are relatively stable. Effective adaptive routing algorithms, which can tune routes for concrete situations, already exist.

Mesh-networks dynamic nature and QoS restrictions (supported services, throughput...) cause a set of specific tasks in projection of network interaction layers. For example one of such a tasks is routes reconfiguration while dynamic mesh node is moving. In opposite to wired networks, where routing take place on the network layer, mesh networks have an ability to use a MAC-layer for that purpose. In dynamic mesh-networks routes reconfiguration happens with high frequency. Thus there is a task of short-time nodes load forecasting for traffic distribution optimization.

Project Summary: 

The main project idea is to study the developing and implementing possibility of an algorithm based on process mining methods aimed to determine template network topologies in dynamic mesh networks. Such an algorithm should trace topology changes and provide an information for routes redistribution among nodes which are supposed to have less traffic load in near future. Current level of understanding of the field of research allows to determine several research directions:

  1. Make a set of dynamic mesh networks topologies in NS-3;
  2. Implement NS-3 plug-in for trace generation in MXML format;
  3. Define a set of process mining algorithms (using ProM framework for process analysis) which are most appropriate for mesh networks;
  4. Propose and implement in NS-3 an algorithm for network hotspots finding in short-time period. Algorithm should be based on process mining techniques;
  5. Develop a metric based on the proposed algorithm and implement it in NS-3
  6. Analyze metric efficiency on a set of dynamic mesh networks topologies
Project goals and future research directions: 
<p class="rtejustify"><em>Main goals:</em></p> <ol> <li class="rtejustify">Investigate routing algorithms and metrics in mesh-networks;</li> <li class="rtejustify">Analyze routing processes using Process Mining methods;</li> <li class="rtejustify">Elaborate an algorithm and method for routes optimization;</li> <li class="rtejustify">Implement an algorithm in a network simulator and try it on several dynamic topologies.</li> </ol> <p class="rtejustify"><em>Future research:</em></p> <ol> <li>Make a set of outdoor experiments and optimize an algorithm and metric for real networks;</li> <li>Implement developed metric in the current mesh-network routing protocol;</li> <li>Integrate metric with QoS service.</li> </ol>
List of team members and their organizations: 

1.    Evgeny Kalishenko, Saint-Petersburg Electrotechnical University “LETI”
2.    Vladislav Saveliev, Academic University of Russian academy of science
3.    Supervisor: Kirill Krinkin, FRUCT LETI Lab

Project Timeline and Expected Deliverables: 
  1. NS-3 plug-in for generating traces in MXML format;
  2. Set of Process Mining algorithms with are most reliable for extracting process templates from ns-3 traces;
  3. An algorithm for short time forecasting of overloaded nodes in dynamic mesh-network;
  4. Routing metric to consider a risk of node overload.
  5. Paper and report on 11th FRUCT seminar.
Final deadline: 
Friday, May 31, 2013 (All day)