Multipath Scheduling for 5G Networks: Evaluation and Outlook

Abstract-The fifth-generation (5G) of cellular networks aims at providing very high data rates, ultra-reliable low latency, and massive connection density. As one of the fundamental design trends toward these objectives, 5G exploits multi-connectivity, i.e., the concurrent use of multiple access networks, where multipath transport protocols have emerged as key technology enablers. The scheduler of a multipath transport protocol determines how to distribute the data packets onto different paths and has a critical impact on the protocol performance. Within such context, this paper presents the first empirical evaluation of state-of-the-art multipath schedulers based on real 5G data, for both static and mobile 5G scenarios. The results show the potential of using multipath transport and the importance of relying on appropriate scheduling algorithms. The reported results further illustrate the benefits of employing a learning-based scheme to automatically design the multipath schedulers for 5G networks, and motivate further studies of advanced learning schemes that can adapt quicker to the path conditions and take into account the emerging features and requirements of 5G and beyond networks.

Supplementary Results

Static scenario: File download throughput for multipath transport between 5G (with different amount of connected users ) and 4G, as a function of the adopted scheduler. Performance for single path transport are reported for comparison.

5G with 1 connected user
5G with 3 connected users
5G with 5 connected users

Static scenario: File download throughput for multipath transport between 5G (with different amount of connected users ) and WiFi, as a function of the adopted scheduler. Performance for single path transport are reported for comparison.

5G with 1 connected user
5G with 3 connected users
5G with 5 connected users

Mobile scenario: File download throughput for multipath transport between 5G mmWave (VZ) and 5G mid-band (Sprint)(with different amount of connected users), as a function of the adopted scheduler. Performance for single path transport are reported for comparison.

5G with 1 connected user
5G with 3 connected users
5G with 5 connected users

In general, the numbers of users connecting to the 5G does not impact the results too much in both scenarios, as the file download task evaluated in this article does not fully utilize all the bandwidth.

Path utilization at different scenarios. The results present some preliminary insights as shown as following. Further insights, however, require the mathematical modelling over multipath transmission, e.g., coupled congestion control, to better understand via the controlled simulation, rather than emulation, how learning based scheduling works in a very high granularity.

While others prioritize lower RTT path, i.e., WLAN, Peekaboo tends to send more packets over 5G
Peekaboo also presents a different type of strategy when the heterogeneity of two paths decrease
Although all performs similarly, Peekaboo tends to send more packets over Sprint

%d bloggers like this: