LBCN: Load Balancing based on Congestion Notification in CRAN Networks for 5G Transport

The progress of modern smart cities has put more requirement on the development of mobile communication, especially the fifth generation (5G) network. Cloud Radio Access Network (CRAN) is an emerging network architecture designed to deal with the bursty and massive 5G traffic. In CRAN networks, where multipath is often the case, an efficient load balancing algorithm is needed to spread load among the links, to provide users promised quality of service (QoS). In the application of 5G communications such as smart cities, the delay of packets and flow completion time (FCT) are highly concerned because QoS is closely allied to these metrics. We propose a load balancing algorithm named LBCN, using two waterlines of buffer occupancy to detect congestion and taking both local and neighbor buffer congestion information into consideration. We simulate LBCN in a 3-tier Clos network. Simulation results shows that our algorithm achieves similar performance to CONGA, shorten the FCT of large flows with low priority, which is a main data source in 5G traffic. It also proves that LBCN balances the utilization of out ports better than ECMP and CONGA.