The modern gaming industry, driven by increasingly immersive and competitive experiences, requires network infrastructures capable of ensuring millisecond-level latency and high Quality of Service (QoS). This thesis analyzes the evolution from traditional data center architectures to distributed cloud-native solutions, focusing on the intelligent orchestration of game servers via Kubernetes and the Agones framework within Edge, IoT, and multi-cloud environments. The research explores the integration of Machine Learning paradigms for proactive resource management. Specifically, it examines persistent storage with dynamic tiering techniques and predictive autoscaling models (based on hybrid Facebook Prophet and LSTM architectures) capable of anticipating traffic spikes, reducing operational costs by 30-40%, and minimizing downtime. A central pillar of this work is energy sustainability. By analyzing standardized KPIs (ISO/IEC 30134) such as PUE, CUE, and WUE, and studying cutting-edge cases like underwater data centers (UDC Lin-gang and Project Natick), the thesis demonstrates how technological innovation can reduce the carbon footprint and improve hardware reliability by up to 8 times compared to terrestrial sites. Finally, the work proposes operational guidelines for Zero Trust security and active monitoring through AIOps, outlining a path toward digital infrastructures that are both technically excellent and environmentally responsible.
L'industria videoludica moderna, spinta verso esperienze sempre più immersive e competitive, richiede infrastrutture di rete capaci di garantire latenze millimetriche e un'elevata qualità del servizio (QoS). Questa tesi analizza l'evoluzione dalle architetture di data center tradizionali verso soluzioni cloud-native distribuite, basate sull'orchestrazione intelligente di game server tramite Kubernetes e il framework Agones in ambienti Edge, IoT e multi-cloud. Il lavoro approfondisce l'integrazione del Machine Learning per la gestione proattiva delle risorse. Nello specifico, viene esaminato lo storage persistente con tecniche di tiering dinamico e modelli di autoscaling predittivo (basati su architetture ibride Facebook Prophet e LSTM) capaci di anticipare i picchi di traffico, riducendo i costi operativi del 30-40% e minimizzando i tempi di inattività. Un pilastro centrale dell'elaborato è la sostenibilità energetica. Attraverso l'analisi dei KPI standardizzati (ISO/IEC 30134) come PUE, CUE e WUE, e lo studio di casi d'eccellenza come i data center sottomarini (UDC Lin-gang e Project Natick), la tesi dimostra come l'innovazione tecnologica possa abbattere l'impronta carbonica e migliorare l'affidabilità hardware fino a 8 volte rispetto ai siti terrestri. Infine, vengono proposte linee guida operative per una sicurezza Zero Trust e un monitoraggio attivo tramite AIOps, delineando un percorso verso infrastrutture digitali tecnicamente eccellenti e ambientalmente responsabili.
Orchestrazione intelligente e sostenibile di game server in ambienti Edge-Cloud: Sicurezza, autoscaling e risparmio energetico.
TRAPANI, FRANCESCO
2024/2025
Abstract
The modern gaming industry, driven by increasingly immersive and competitive experiences, requires network infrastructures capable of ensuring millisecond-level latency and high Quality of Service (QoS). This thesis analyzes the evolution from traditional data center architectures to distributed cloud-native solutions, focusing on the intelligent orchestration of game servers via Kubernetes and the Agones framework within Edge, IoT, and multi-cloud environments. The research explores the integration of Machine Learning paradigms for proactive resource management. Specifically, it examines persistent storage with dynamic tiering techniques and predictive autoscaling models (based on hybrid Facebook Prophet and LSTM architectures) capable of anticipating traffic spikes, reducing operational costs by 30-40%, and minimizing downtime. A central pillar of this work is energy sustainability. By analyzing standardized KPIs (ISO/IEC 30134) such as PUE, CUE, and WUE, and studying cutting-edge cases like underwater data centers (UDC Lin-gang and Project Natick), the thesis demonstrates how technological innovation can reduce the carbon footprint and improve hardware reliability by up to 8 times compared to terrestrial sites. Finally, the work proposes operational guidelines for Zero Trust security and active monitoring through AIOps, outlining a path toward digital infrastructures that are both technically excellent and environmentally responsible.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi Francesco Trapani.pdf
accesso aperto
Descrizione: La tesi analizza l'integrazione di tecnologie cloud-native e modelli di Machine Learning per ottimizzare l'infrastruttura di rete dedicata al gaming multiplayer.
Dimensione
3.34 MB
Formato
Adobe PDF
|
3.34 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/25661