AHP-based multi-criteria analysis of multi-cloud data management techniques

Abstract

Today, the multi-cloud concept covers more and more spheres of modern life: tech-related industries, financial services, healthcare, etc. A multi-cloud environment combines services from different providers through a specific user-side architecture. The aviation industry has several characteristics: large-scale modeling, testing, and managing vast data. Running complex simulations is time-consuming and requires significant high-performance computing (HPC) resources. This, in turn, makes using multi-cloud very promising in this industry. However, currently, there are many approaches to multi-cloud interaction, each of which has its own characteristics. The article researches different approaches to multi-cloud data access and create a model to determine the most optimal one. The research subject is interaction methods in multi-cloud systems: Multi-cloud data storage gateways using the example of S3Proxy, data management platforms using the example of Apache NiFi, and cloud-agnostic libraries using the example of Apache Libcloud. Their main advantages, disadvantages, and features of use are given. The research tasks are formalizing the problem, defining cost, performance, security, and implementation effort parameters for each approach, and developing a multi-criteria decision analysis (MCDA) model using the Analytical Hierarchy Process (AHP) method. Thanks to its adaptability, this model allows organizations to choose the most effective strategy for integrating multi-cloud technologies into their work processes, maximizing potential benefits, regardless of the specific context. The following results were obtained. The MCDA/AHP model was built with the input parameters of performance, security, storage cost, and implementation effort determined. In this case, performance was found empirically, cost was based on the AWS S3 pricing model, implementation efforts were estimated based on expert opinion, and the security criterion was determined using the weighted scoring method. It is important to note that the cost, performance, security, and implementation complexity criteria are ranked in descending order of importance and play a crucial role in obtaining the initial values. Cloud-agnostic libraries achieved the best results, followed by multi-cloud storage gateways and data management platforms. Conclusions. The scientific novelty of this work is the development of a multi-criteria model for determining the most optimal multi-cloud approach. The limitations and opportunities of MCDA/AHP are also described. This not only helps to determine the best approach for specific requirements but also lays a solid foundation for further research and development of strategies for the use of multi-cloud environments in various industries, paving the way for future advancements in the field.