Hierarchical and Density Aware Spectral Clustering Approach to Server Placement for Collaborative Virtual Services

Abstract

Boundaries on separate infrastructure domains are blurring as infrastructure providers are collaborating to offer end-to-end services. To capture opportunities and shine in competitive service market, infrastructure and service providers need to excel in addressing the changing customer requirements and in the operations and management of resources. They should welcome effective collaboration with other network and infrastructure providers in delivering quality services to customers. Effective sharing of the infrastructure resources is essential in meeting the customer demands and reducing the cost. Server placement problem for end-to-end virtual services becomes a crucial optimization challenge for providers in such collaborative environments. In our previous work, we formulated the collaborative virtual server placement problem and suggested density-based clustering algorithms to address this problem. In this work, we propose hierarchical and density-aware spectral clustering as a feasible solution to the server placement problem in virtual and collaborative service environments. We introduce three enhancements to the classical spectral clustering to make it fit to our problem: vertex weights, density awareness, hierarchical partitioning.