This thesis studies the solutions to improve user quality of experience to cope with the exponential traffic growth through aggregate LTE and Wi-Fi resources in radio access network and recycling contents with cache in core oredge networks. Firstly, we consider a smart aggregation of LTE and Wi-Fi radios for resource scarcity of mobilecellular networks, which is called LTE-W that efficiently utilizes LTE and Wi-Fi links (link-level aggregation) onlywith the minimum changes of eNodeBs, LTE backhaul networks, and mobile terminals, thus easily deployable.LTE-W has the two key components such as mode selection to consider intra-cell fairness rather than just following users' intention of aggregation and bearer-split scheduling into LTE and Wi-Fi links for the performance of TCP flows. Secondly, we asymptotically study the average delay performance for random content requests in a generalnetwork of caches under on-line repeated learning and content placement policy in terms of three aspects suchas how long we should learn popularity, how often we should change cached contents, and how we use learntpopularity in caching contents over the network under various content characters of content popularity and averagerouting distance in network of caches.