Improving file management in a digital construction platform

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Electrical Engineering | Master's thesis

Department

Mcode

Language

en

Pages

44

Series

Abstract

The construction industry has been slowly undergoing a digital transformation, part of which is the digitalization of files generated during the lifecycle of a construction project. These files can be anything from simple text files to complicated digital designs of buildings. Efficient management and processing of diverse project management files are important in the construction industry because real-time data access is required. Therefore, it is important that the digital construction platforms that handle these files also have a scalable and resilient underlying architecture. Infrakit is a Finnish digital construction platform that serves infrastructure projects. However, the platform faces bottlenecks in file processing related to scalability, latency, and resource utilization, which can affect Infrakit customers. This thesis investigates the limitations of the legacy file processing system of Infrakit. It was observed that Infrakit’s existing monolithic architecture uses a work queue system with constrained threadpools and block storage. This results in processing latency and limited scalability for heavy workloads. To overcome these bottlenecks, a new event-driven architecture with serverless components, such as AWS Lambda, AWS Batch, and scalable object storage (AWS S3), was proposed and implemented in an experimental setup. An empirical benchmarking study was then conducted on both the legacy and new architectures using synthetic workloads that closely resembled real-world project management files. Several metrics were collected, such as the overall processing time, parsing time, memory usage, and queue wait time, which were further analyzed. The results demonstrate that the new architecture eliminates the bottlenecks of the existing legacy system by eliminating the queue waiting time by 100%. Scalability was improved by offloading the computation to serverless services, resulting in an 85% reduction in processing latency time. Finally, the memory usage in the main application was reduced by approximately 47%. Overall, the proposed architecture provides a scalable, resilient, and robust environment for processing project management files.

Description

Supervisor

Premsankar, Gopika

Thesis advisor

Hokkanen, Visa
Toivonen, Kim

Other note

Citation