A Scalable Location Service Supporting Overload Situations

Luis Bernardo                                                                                                     Paulo Pinto
Faculdade de Ciências e Tecnologia                                                              Faculdade de Ciências e Tecnologia
Universidade Nova de Lisboa                                                                          Universidade Nova de Lisboa
Inesc, R. Alves Redol, 9 P-1000 Lisboa Portugal                                                   P-2825 Monte da Caparica, Portugal
lflb@inesc.pt                                                                                                   pfp@uninova.pt



Abstract.

This paper addresses the issue of defining a location service suitable for very dynamic and highly populated networks (millions of users), where services might experience highly correlated peaks of traffic or synchronized access to specific servers. A cooperative mobile agent solution is proposed to solve the major problems, allowing the dynamic deployment of new application servers when needed. But it requires an adequate location service to route the clients to the application servers, which scales to a large number of clients and still allows a high number of updates. This paper presents a very dynamic location service, which adapts to overload situations by modifying the routing information distribution and (possibly) its servers’ internal structure. The responsive and fast route update results in a load re-distribution, which allows it to scale to a broader range compared to other alternative approaches.
 

To appear in:

Third International Workshop on Artificial Intelligence in Distributed Information Networking (AiDIN'99), Orlando, USA, July 18, 1999, Workshop Notes