Comparative Performance Analysis of Desktop and Cloud-Based Architectures for Network Dimensioning Tools : A Case Study at Ericsson AB

Sammanfattning

Network dimensioning tools play a critical role in telecommunications infrastructure planning, yet many remain deployed as legacy desktop applications constrained by local hardware limitations. This study presents a comparative performance analysis of desktop and cloud-based architectures for CANDI IMS SUPERMAN, an internal Ericsson tool used for IMS network dimensioning and capacity planning. A containerized web-based prototype was developed using FastAPI and Docker, rep- licating the core dimensioning logic of the legacy application while introducing a browser-based interface and parallel multiprocessing capabilities. Performance benchmarking was conducted across three deployment environments—the legacy desktop application, a locally containerized Docker instance, and a remote Ericsson private cloud server—using four datasets ranging from 57.5 MB to 35.4 GB, totaling 60 controlled runs. For small datasets, the legacy desktop and cloud-based prototype perform compara- bly. For larger datasets, the cloud architecture offers significant advantages—con- tainerization alone on identical hardware reduced execution time for the largest da- taset by 43% in single-core mode. With multi-core parallelization on the remote server, execution time was further reduced from 2,280 seconds to 382 seconds, an 83% reduction, while system-wide CPU utilization remained well below server ca- pacity, indicating substantial headroom for further scaling. Beyond the empirical findings, the multi-core results had a direct organizational im- pact—the product owner formally created a development backlog item to integrate multi-core processing into the legacy desktop application upon presentation of the results. The findings demonstrate that cloud-based deployment is particularly beneficial for computationally intensive workloads while introducing negligible overhead for smaller datasets. This provides empirical evidence to support architectural transi- tion decisions for specialized telecommunications dimensioning tools.