Network control and management techniques (e.g., dynamic path switching, on-demand bandwidth provisioning) rely on active measurements of end-to-end network status. The measurements are needed to meet network monitoring objectives such as: (a) intra-domain/inter-domain paths status checking, (b) accurate network weather forecasting, (c) accurate Service Level Agreement (SLA) compliance, and (d) rapid anomaly event detection. Recent widespread deployment of openly accessible multi domain active measurement frameworks such as perfSONAR has resulted in users competing for system and network measurement resources. Since the measurement
resources (i.e., tool servers, network bandwidth) are limited, it might not be possible for a measurement scheduler to accommodate all user requests i.e., generate completely feasible schedules under high measurement request loads. Consequently, measurement requests that could not be scheduled might adversely affect monitoring
accuracy needed in critical resource adaptation decisions. Moreover, lack of semantic priorities might block intra-domain measurement requests that are more important than inter-domain measurement requests from an ISP’s perspective. In such cases, there is a need to prioritize measurement requests by using semantic priorities based on user and resource policies specified by the ISP’s and also to prevent resource contention caused by limited measurement resources.
In this thesis, we present a novel ontology based semantic priority scheduling (SPS) algorithm that handles resource contention while servicing measurement requests for meeting network monitoring objectives that aid in resource adaptation decisions within multi-domain measurement federations. We adopt ontologies to formalize
semantic definitions and developed an inference engine to dynamically prioritize measurement requests. Performance evaluation results demonstrate that our SPS algorithm outperforms existing deterministic and heuristic algorithms in terms of user ’satisfaction ratio’ and ’average stretch’ amongst serviced measurement requests. Further, using sampling experiments on real-network perfSONAR measurement data sets, we show that our SPS algorithm successfully mitigates oversampling and further improves satisfaction ratio. Our SPS scheme and evaluation results are vital to manage large-scale measurement infrastructures used for meeting monitoring objectives in next-generation applications and networks.