In recent years, the Internet traffic explosion together with the deployment of next-generation Internet services, such as online gaming, mission critical communications, virtual reality, etc., has posed serious challenges to network operators, who need to find solutions to cope with high-capacity requests and strict QoS (Quality of Service) requirements. In order to provide larger capacity to end users, the number of Base Stations will grow dramatically, thus leading to cost and energy efficiency issues, especially in the access/aggregation network. To solve these issues, a new access/aggregation network architecture was proposed, referred to as Centralized Radio Access Network (C-RAN). In this architecture, the Remote Radio Head (RRH) and the Digital Unit (DU) of each Base Station(BS) are no more co-located at the cell site, as in the traditional RAN architecture, but they are located separately. In particular, the RRH remains at the cell site, while the DU can be place in other centralized locations. The split of the RRH and the DU results in a new type of traffic to be handled by the network, i.e., the fronthaul traffic. This type of traffic has strict latency and capacity requirements which need to be taken into consideration in the design and operation of the aggregation network. Due to the high capacity requirements, the most adapted fronthaul transport solution is based on optical fibers and on wavelength-division multiplexing (WDM), which allows to multiplex multiple signals over the same fiber. In this study, we address the "Dynamic Handover of BaseBand Processing in a WDM access/aggregation network" problem under a dynamic traffic scenario. In particular, our proposal consists of dynamically allocating network and BaseBand processing resources according to users traffic. In detail, our novel solution is based on the implementation of the DU handover, allowing RRHs to change their associated DUs, based on traffic load. We have developed four algorithms in order to evaluate the impact of DU handover, traffic grooming, traffic bifurcation and amount of DU handover on performances. Another key concept of our work is the use of functional splits in order to reduce the high bandwidth request imposed by fronthaul traffic. By splitting the baseband processing chain in a different point with respect to the classical fronthauling solution, the traffic exchanged between the RRH and the DU pool is named midhaul traffic. Since this traffic requires less bandwidth than fronthaul, this new approach is nowadays widely studied and seen as a promising solution to aggregate mobile-native traffic in metro/access networks, while still allowing sharing of processing resources. To obtain practical results, a discrete-event network simulator has been used in order to test the developed algorithms and to compare performances, considering a network topology based on a WDM access/aggregation network. From the results obtained, it emerges that the algorithm enabling the unconstrained DU handover, traffic grooming and traffic bifurcation is the best in terms of blocking probability and consolidation of resources. In particular, for the considered case studies, a reduction of up to 55\% in the normalized average number of active pools, for medium loads, is provided in comparison to existing algorithms where DU handover is not performed. For high loads, a reduction up to 20\% is achieved. Moreover, substantial improvements are provided also in terms of blocking probability, which is reduced to zero for all the traffic loads. By comparing the four algorithms we have developed, a reduction of up to 35\% in the normalized average number of active pools is provided by the most flexible algorithm with respect to other three algorithms, for medium and high loads. In addiction, considerable improvements are achieved in terms of blocking probability by the three algorithms with respect to the one where grooming is disable, especially for low and medium loads.
Negli ultimi anni, l'esplosione del traffico Internet insieme alla diffusione di servizi Internet di nuova generazione, come online gaming, comunicazioni mission-critical, realtà virtuale, ecc., ha creato sfide difficili per gli operatori, che sono in cerca di soluzioni per soddisfare le richieste di alta capacità e requisiti di QoS (Quality of Service). Per garantire una maggiore capacità agli utenti finali, il numero di stazioni radio base aumenterà drasticamente, causando problemi di costo e efficienza energetica, soprattutto nella rete di accesso e aggregazione. Per risolvere questi problemi, una nuova architettura per la rete di accesso e aggregazione è stata proposta, alla quale ci si riferisce con il termine Centralized Radio Access Network (C-RAN). In questa architettura, Remote Radio Head (RRH) e Digital Unit (DU) non sono più collocate insieme nella cella, come avviene nell'architettura RAN tradizionale, ma sono posizionate separatamente. In particolare, l'RRH rimane posizionata alla cella, mentre la DU può essere posizionata in altri luoghi centralizzati. La separazione tra RRH e DU comporta l'instradamento di un nuovo tipo di traffico, che deve essere gestito dalla rete, detto traffico di fronthaul. Questo tipo di traffico ha requisiti di bassa latenza e alta capacità che devono essere tenuti in considerazione nella progettazione e nel funzionamento della rete di aggregazione. Proprio per questo motivo, la soluzione più adatta per il trasporto del traffico di fronthaul si basa su fibra ottica e multiplazione a divisione di lunghezza d'onda (WDM), che consente di multiplare più segnali su una stessa fibra. In questa tesi, analizziamo il problema di "Dynamic Handover of BaseBand Processing in a WDM access/aggregation network", considerando uno scenario di traffico dinamico. In particolare, la nostra proposta consiste nell'allocare dinamicamente le risorse di processing in banda base a seconda del traffico generato dagli utenti. Nel dettaglio, la soluzione proposta si basa sull'implementazione del cambio di DU (DU handover), che consente alle unità RRH di cambiare la loro DU associata, a seconda del carico offerto. Abbiamo sviluppato quattro algoritmi per valutare l'impatto sulle performance del DU handover, dell'aggregazione e birforcazione del traffico e del DU handover con vincolo temporale. Un altro concetto chiave alla base del nostro lavoro è l'utilizzo dei cosiddetti "functional split", che hanno l'intento di ridurre i requisiti di alta capacità imposti dal traffico di fronthaul. Separando la catena delle funzioni di processing in banda base in diversi punti rispetto alla classica soluzione di fronthaul, il traffico scambiato tra la RRH e il DU pool prende il nome di traffico di midhaul. Dato che questo traffico richiede meno capacità del fronthaul, questo nuovo approccio è al giorno d'oggi ampiamente studiato e percepito come una soluzione promettente per aggregare il traffico di tipo mobile nelle reti metro/access, sempre consentendo la condivisione delle risorse. Per ottenere risultati pratici, è stato sviluppato un simulatore di rete a eventi discreti, che è stato poi utilizzato per testare gli algoritmi proposti e confrontarne le performance, considerando una topologia di rete di accesso e aggregazione basata su WDM. Dai risultati ottenuti, è emerso che l'algoritmo che permette il cambio della DU senza vincoli, che consente l'aggregazione e la biforcazione del traffico è il migliore in termini di probabilità di blocco e consolidamento delle risorse. In particolare, per i casi studio considerati, si ottiene una riduzione fino 55\% nel numero normalizzato di DU pools medi attivi rispetto all'algoritmo in cui il DU handover non è consentito, per carichi medi. Per carichi alti, si raggiunge una riduzione fino al 20\%. In più, importanti miglioramenti sono raggiunti in termini di probabilità di blocco, che è ridotta a zero per tutti i carichi considerati. Confrontando i quattro algoritmi che abbiamo sviluppato, una riduzione fino al 35\% nel numero normalizzato di nodi attivi medi si ottiene con l'algoritmo più flessible rispetto agli altri tre algoritmi, per carichi medi e alti. Significanti miglioramenti sono ottenuti in termini di probabilità di blocco dai tre algoritmi rispetto all'algoritmo in cui il grooming non è consentito, soprattutto per carichi bassi e medi.
Dynamic handover of baseband processing in a WDM access-aggregation network
MAGONI, MONICA
2017/2018
Abstract
In recent years, the Internet traffic explosion together with the deployment of next-generation Internet services, such as online gaming, mission critical communications, virtual reality, etc., has posed serious challenges to network operators, who need to find solutions to cope with high-capacity requests and strict QoS (Quality of Service) requirements. In order to provide larger capacity to end users, the number of Base Stations will grow dramatically, thus leading to cost and energy efficiency issues, especially in the access/aggregation network. To solve these issues, a new access/aggregation network architecture was proposed, referred to as Centralized Radio Access Network (C-RAN). In this architecture, the Remote Radio Head (RRH) and the Digital Unit (DU) of each Base Station(BS) are no more co-located at the cell site, as in the traditional RAN architecture, but they are located separately. In particular, the RRH remains at the cell site, while the DU can be place in other centralized locations. The split of the RRH and the DU results in a new type of traffic to be handled by the network, i.e., the fronthaul traffic. This type of traffic has strict latency and capacity requirements which need to be taken into consideration in the design and operation of the aggregation network. Due to the high capacity requirements, the most adapted fronthaul transport solution is based on optical fibers and on wavelength-division multiplexing (WDM), which allows to multiplex multiple signals over the same fiber. In this study, we address the "Dynamic Handover of BaseBand Processing in a WDM access/aggregation network" problem under a dynamic traffic scenario. In particular, our proposal consists of dynamically allocating network and BaseBand processing resources according to users traffic. In detail, our novel solution is based on the implementation of the DU handover, allowing RRHs to change their associated DUs, based on traffic load. We have developed four algorithms in order to evaluate the impact of DU handover, traffic grooming, traffic bifurcation and amount of DU handover on performances. Another key concept of our work is the use of functional splits in order to reduce the high bandwidth request imposed by fronthaul traffic. By splitting the baseband processing chain in a different point with respect to the classical fronthauling solution, the traffic exchanged between the RRH and the DU pool is named midhaul traffic. Since this traffic requires less bandwidth than fronthaul, this new approach is nowadays widely studied and seen as a promising solution to aggregate mobile-native traffic in metro/access networks, while still allowing sharing of processing resources. To obtain practical results, a discrete-event network simulator has been used in order to test the developed algorithms and to compare performances, considering a network topology based on a WDM access/aggregation network. From the results obtained, it emerges that the algorithm enabling the unconstrained DU handover, traffic grooming and traffic bifurcation is the best in terms of blocking probability and consolidation of resources. In particular, for the considered case studies, a reduction of up to 55\% in the normalized average number of active pools, for medium loads, is provided in comparison to existing algorithms where DU handover is not performed. For high loads, a reduction up to 20\% is achieved. Moreover, substantial improvements are provided also in terms of blocking probability, which is reduced to zero for all the traffic loads. By comparing the four algorithms we have developed, a reduction of up to 35\% in the normalized average number of active pools is provided by the most flexible algorithm with respect to other three algorithms, for medium and high loads. In addiction, considerable improvements are achieved in terms of blocking probability by the three algorithms with respect to the one where grooming is disable, especially for low and medium loads.| File | Dimensione | Formato | |
|---|---|---|---|
|
tesi_depositata2.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Tesi di laurea
Dimensione
2.17 MB
Formato
Adobe PDF
|
2.17 MB | 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/10589/141904