Full-Duplex vs. Half-Duplex: Delivery-Time Optimization in Cellular Downlink
Conference: European Wireless 2017 - 23th European Wireless Conference
05/17/2017 - 05/19/2017 at Dresden, Germany
Proceedings: European Wireless 2017
Pages: 6Language: englishTyp: PDFPersonal VDE Members are entitled to a 10% discount on this title
Kariminezhad, Ali; Gherekhloo, Soheil; Sezgin, Aydin (Ruhr-Universität Bochum, Germany)
The employment of the emerging trend of proactive caching leads to a trade-off between memory size and information rate. Due to the limited cache size only a limited number of demanded files might be available in the cache of the local server. Hence, the missing files need to be forwarded to the destinations. Now, the local server has the option to operate in either half-duplex or full-duplex mode. In half-duplex mode (HD), the demanded files are transmitted as soon as all files are provided to the local server. However, in a full-duplex mode (FD), the cached files can be communicated to the users, meanwhile receiving the missing files. Hence, by full-duplex operation, the local servers manage their cache (deplete/refill) in real-time adaptively which is beneficial from memory perspective. For both cases, i.e., HD and FD, the local server exploits decodeand- forward (DF) strategy. In this paper, we address these two strategies from the delivery-time perspective. As the worst link is the delivery-time bottleneck, fairness among users becomes of particular interest. Hence, we cast this problem as a min-max fair optimization problem. Moreover, we formulate the problem as a semi-definite relaxation (SDR) feasibility check problem after defining auxiliary variables to cope with non-convex constraints. The SDR comes along with bisection (over a single variable) and exhaustive search (over a few variables). Eventually, we compare the performance of half-duplex and full-duplex operations from the delivery-time perspective. Depending on the self-interference (SI) channel strength, the full-duplex operation can be beneficial from the delivery-time perspective up to almost 10% compared to the half-duplex counterpart.