We are facing an increasing interest about virtualization technologies that enable new paradigms and technical approaches like Cloud Computing, Software Defined Networks. Infrastructure Providers (InPs), managing distributed physical resources, need to perform the mapping of the requested virtual resources on the physical ones. For this purpose, they need to optimize the allocation process in order to: (1) optimally use the various types of available resources shared among customers, and (2) be compliant with Service Level Agreement (SLA) to meet the QoS attributes. This mapping phase, called Virtual Network Embedding, is known to be an NP-hard problem. The aim of the PhD thesis project described in this paper is to propose algorithmic solutions that allow the InP to optimize the mapping of virtual computational nodes and network links in order to improve objective metrics (e.g., revenue) while satisfying multiple QoS attributes agreed in the SLA. To achieve this goal, the thesis will address these main topics: (1) efficient online heuristics for network-aware embedding of multiple network requests, (2) multi-constrained embedding according multiple QoS metrics (e.g., maximum expected delay). Early results show that our algorithms performs better than existent ones, encouraging us to continue with the future work.