Evaluating RPC for Cloud-Native 5G Mobile Network Applications

Abstract

This thesis investigates the communication between services in 5G network functions. The development of the 5G Core (5GC) is by design increasing the amount of communication needed in the control plane. The reason for this is the migration to the cloud and the adoption of a microservices architecture. The telecommunications domain sets strict requirements on performance, which implies the need for the implementation of inter-service communication to be carefully constructed. This thesis evaluates the use of Remote Procedure Call (RPC) as inter-service communication in a 5GC network function. The purpose is to evaluate whether RPC frameworks will fulfill the requirements of inter-service communication and the strict requirements on telecom applications. The frameworks evaluated are gRPC and Apache Thrift. We also compare the frameworks to a TCP solution since this is the approach currently considered for this use case and a solution with minimal overhead to the communication. The evaluation is both quantitative, with benchmarks on latency, throughput and CPU usage, and qualitative where qualities such as availability and ease of development are evaluated. From the evaluation, we can conclude that using RPC frameworks would suit most needs. Even if the evaluated RPC frameworks perform slightly worse than a reference TCP solution in the quantitative evaluation, they can provide many other benefits such as bidirectional streaming RPC and highavailability features. Among the evaluated RPC frameworks, Apache Thrift stands out slightly in terms of performance, while gRPC stands out in the qualitative evaluation.