In the modern era of ubiquitous digital services, microservice architectures have become the standard for building scalable large-scale systems. However, traditional inter-service communication relies on direct protocols that necessitate complex service discovery mechanisms, leading to tight architectural coupling and increased sensitivity to network topology. This thesis proposes an architectural framework designed to overcome these limitations by establishing a broker-mediated communication substrate that prioritizes systemwide resilience and operational autonomy. By leveraging the Advanced Message Queuing Protocol (AMQP), the architecture achieves complete location transparency, effectively decoupling business logic from the underlying infrastructure. This shift enables inherent fault tolerance, as the system natively buffers requests and manages transient failures without data loss or service interruptions. A core contribution of the proposed solution is an intelligent self-managing layer that orchestrates the system’s lifecycle across geographically distributed clusters. This layer features an adaptive scaling strategy that autonomously manages service presence in multiple regions. By analyzing real-time traffic demand and queue dynamics, the system independently decides when and where to instantiate new components, effectively balancing the trade-off between performance and infrastructure costs without manual intervention. Experimental evaluations demonstrate that this broker-mediated approach provides a robust foundation for building self-healing distributed systems. The results confirm that the proposed self-managing logic effectively minimizes resource waste while maintaining a performance profile highly competitive with traditional models. Ultimately, this work provides a scalable and cost-effective framework capable of adapting to fluctuating global demand while hiding the complexity of multiregion deployments from the application layer.
Nell’attuale era in cui i sistemi informativi sono onnipresenti, le architetture a microservizi sono diventate lo standard per la costruzione di sistemi scalabili. Tuttavia, la comunicazione tra servizi si affida comunemente a protocolli diretti che necessitano di complessi meccanismi di service discovery, comportando una forte dipendenza architetturale e una elevata sensibilità alla topologia di rete. Questa tesi propone un framework progettato per superare tali limitazioni attraverso la creazione di un canale comunicativo mediato da broker che pone come priorità la resilienza dell’intero sistema e l’autonomia operativa. Sfruttando il protocollo AMQP (Advanced Message Queuing Protocol), l’architettura garantisce una completa comunicazione trasparente all’indirizzo di rete, disaccoppiando efficacemente la logica di business dall’infrastruttura sottostante. Questo approccio abilita una tolleranza ai guasti intrinseca, poiché il sistema è in grado di gestire nativamente il buffering delle richieste e i guasti temporanei senza perdita di dati o interruzioni del servizio per l’utente finale. Un contributo centrale della soluzione proposta è un sistema intelligente di autogestione che, attraverso una strategia di scaling adattivo, orchestra il ciclo di vita dei servizi tra cluster distribuiti geograficamente. Analizzando in tempo reale la domanda di traffico, il sistema decide in modo indipendente quando e dove istanziare nuovi componenti, bilanciando efficacemente il compromesso tra prestazioni e costi infrastrutturali senza necessità di intervento manuale. Le valutazioni sperimentali dimostrano che questo approccio mediato da broker fornisce una base robusta per la costruzione di sistemi distribuiti auto-gestiti. I risultati confermano che la logica di autogestione proposta minimizza efficacemente lo spreco di risorse, mantenendo al contempo un profilo prestazionale altamente competitivo rispetto ai modelli tradizionali. In conclusione, questo lavoro fornisce un framework scalabile ed efficiente, capace di adattarsi a una domanda globale fluttuante e di nascondere la complessità delle distribuzioni multi-regione allo strato applicativo.
Improving the resilience and flexibility of microservice architecture
Bonilla Quiroz, Andres
2025/2026
Abstract
In the modern era of ubiquitous digital services, microservice architectures have become the standard for building scalable large-scale systems. However, traditional inter-service communication relies on direct protocols that necessitate complex service discovery mechanisms, leading to tight architectural coupling and increased sensitivity to network topology. This thesis proposes an architectural framework designed to overcome these limitations by establishing a broker-mediated communication substrate that prioritizes systemwide resilience and operational autonomy. By leveraging the Advanced Message Queuing Protocol (AMQP), the architecture achieves complete location transparency, effectively decoupling business logic from the underlying infrastructure. This shift enables inherent fault tolerance, as the system natively buffers requests and manages transient failures without data loss or service interruptions. A core contribution of the proposed solution is an intelligent self-managing layer that orchestrates the system’s lifecycle across geographically distributed clusters. This layer features an adaptive scaling strategy that autonomously manages service presence in multiple regions. By analyzing real-time traffic demand and queue dynamics, the system independently decides when and where to instantiate new components, effectively balancing the trade-off between performance and infrastructure costs without manual intervention. Experimental evaluations demonstrate that this broker-mediated approach provides a robust foundation for building self-healing distributed systems. The results confirm that the proposed self-managing logic effectively minimizes resource waste while maintaining a performance profile highly competitive with traditional models. Ultimately, this work provides a scalable and cost-effective framework capable of adapting to fluctuating global demand while hiding the complexity of multiregion deployments from the application layer.| File | Dimensione | Formato | |
|---|---|---|---|
|
Polimi___Nova___Master_Thesis_Extended.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Extended version
Dimensione
1.52 MB
Formato
Adobe PDF
|
1.52 MB | Adobe PDF | Visualizza/Apri |
|
Polimi___Nova___Master_Thesis_Executive_Summary.pdf
accessibile in internet per tutti
Descrizione: Executive Summary version
Dimensione
756.58 kB
Formato
Adobe PDF
|
756.58 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/253470