Self-Management Principles in Autonomic Service Architecture Supported by Load Balancing Algorithm

Abstract

This paper is about the role of a heuristic algorithm for load balancing that can be incorporated in Multi-Protocol Label Switching/ Differentiated Services (MPLS/DS) networks based on self-management principles for automated traffic configuration. Delivery of Quality of Services (QoS) differentiation according to specific service level agreements, Service Level Agreement (SLA), has to be in relation to resource management and bandwidth allocation, to ensure better end-to-end QoS provisioning and to avoid traffic congestion. To ensure simultaneous traffic flows with different priorities we need some load-balancing procedure. It could start much earlier than usual, possibly during SLA negotiation. If SLA creation is supported by load control mechanism it can ensure better performances of the network in the moment of service invocation. LSP (Label Switching Path) creation could be influenced by former contracted SLAs and such approach could be more effective than pure routing technique based on the shortest path algorithm (LSP creation in the moment of service invocation). So we propose application of heuristic algorithm tested on numerical examples with maximum M routers on the path and with differentiation of N service classes. Further, we introduced some capacity state restrictions in the process of network optimization, considering different algorithm options. We compared their performances, especially the algorithm complexity that is very important for efficient load control in huge networks.