The advent of new technologies and services with stringent requirements of bandwidth and latency has led to a downward curve in the revenues of telecom operators in the recent years. This has resulted in a paradigm shift, from the deployment of hardware resources to solve the issue at hand, to placement of Virtual Network Functions (VNFs) in the form of virtual machines on low-cost servers and switches in order to reduce expenditures. The VNFs forming the backbone of the Network Function Virtualization Infrastructure (NFVI), are chained together and placed optimally in accordance to the QoS requirements of the services offered, also known as Service Function Chain (SCF) to add flexibility in the existing infrastructure. However, the placement of the VNFs with minimal activation of VNFs and low blocking probability requires an efficient algorithm to be deployed. Moreover, the algorithm also needs to address the challenges brought about by VNF deployment, including maintaining network performance and latency, added security threats, compatibility among physical and virtual entities and manageability of the function among others. Previous works in existing literature have addressed the above problem by proposing various algorithms in a static context, where the number of users, traffic load and network services are considered to be fixed. Existing literature also highlights algorithms in a dynamic context with variable user flows in terms of VNF type and location traversed and varying user types in the network. However, there is a gap in addressing the cases where the type of SC requested is variable with varying arrival rates of users and active user requests. We address this gap in the existing literature by proposing an algorithm which performs dynamic VNF placement considering the SC requested by the users while minimizing the number of VNF instances as well as the blocking probability. Moreover, the algorithm also considers bandwidth constraint of links, computational capacity of the NFV-nodes and latency requirements of the requested SC. We propose four versions of the algorithm with varying levels of SC-awareness and the presence or absence of Latency Island concept. We simulate the different versions of our algorithm in the context of a 4-stage aggregation metro network for remotely and locally served SC cases. We illustrate the results in terms of blocking probability and average number of activated VNFs.
L’avvento di nuove tecnologie e servizi con requisiti rigorosi di larghezza di banda e latenza ha portato ad una curva in discesa nei ricavi degli operatori di telecomunicazioni negli ultimi anni. Ciò ha portato a un cambiamento di paradigma, dalla distribuzione di risorse hardware per risolvere il problema in questione, al posizionamento di funzioni di rete virtuale (VNF) sotto forma di macchine virtuali su server e switch a basso costo al fine di ridurre le spese. I VNF che costituiscono la spina dorsale dell’infrastruttura di virtualizzazione di rete (NFVI), sono incatenati e posizionati in modo ottimale in conformità ai requisiti QoS dei servizi offerti, noti anche come Service Function Chain (SCF) per aggiungere flessibilità nell’infrastruttura esistente. Tuttavia, il posizionamento dei VNF con minima attivazione di VNF e bassa probabilità di blocco richiede un efficiente algoritmo da implementare. Inoltre, l’algoritmo deve anche affrontare le sfide portate avanti dal dispiegamento di VNF, incluso il mantenimento delle prestazioni e della latenza della rete, aggiunte di minacce alla sicurezza, compatibilità tra le entità fisiche e virtuali e la gestione della funzione tra gli altri. I precedenti lavori nella letteratura esistente hanno affrontato il problema di cui sopra suggerendo diversi algoritmi in un contesto statico, in cui il numero di utenti, il carico di traffico ei servizi di rete sono considerati fissi. La letteratura esistente evidenzia anche algoritmi in un contest dinamico con flussi utente variabili in termini di tipo e posizione VNF attraversati e vari tipi di utente nella rete. Tuttavia, vi è un divario nell’affrontare i casi in cui il tipo di SC richiesto è variabile con tassi di arrivo variabili di utenti e richieste di utenti attivi. Facciamo questo divario nella letteratura esistente proponendo un algoritmo che esegue un posizionamento dinamico di VNF considerando la SC richiesta dagli utenti riducendo al minimo il numero di istanze VNF così come la probabilità di blocco. Inoltre, l’algoritmo considera anche il vincolo di larghezza di banda dei collegamenti, la capacità computazionale dei nodi NFV e i requisiti di latenza del SC richiesto. Proponiamo quattro versioni dell’algoritmo con diversi livelli di consapevolezza della SC e la presenza o assenza del concetto di Latency Island. Simulamo le diverse versioni del nostro algoritmo nel contesto di una rete metropolitana di aggregazione a 4 stadi per i casi SC in remoto e localmente serviti. Illustriamo i risultati in termini di probabilità di blocco e numero medio di VNF attivati.
Virtual network function placement for dynamic service chain provisioning
ASKARI, LEILA
2016/2017
Abstract
The advent of new technologies and services with stringent requirements of bandwidth and latency has led to a downward curve in the revenues of telecom operators in the recent years. This has resulted in a paradigm shift, from the deployment of hardware resources to solve the issue at hand, to placement of Virtual Network Functions (VNFs) in the form of virtual machines on low-cost servers and switches in order to reduce expenditures. The VNFs forming the backbone of the Network Function Virtualization Infrastructure (NFVI), are chained together and placed optimally in accordance to the QoS requirements of the services offered, also known as Service Function Chain (SCF) to add flexibility in the existing infrastructure. However, the placement of the VNFs with minimal activation of VNFs and low blocking probability requires an efficient algorithm to be deployed. Moreover, the algorithm also needs to address the challenges brought about by VNF deployment, including maintaining network performance and latency, added security threats, compatibility among physical and virtual entities and manageability of the function among others. Previous works in existing literature have addressed the above problem by proposing various algorithms in a static context, where the number of users, traffic load and network services are considered to be fixed. Existing literature also highlights algorithms in a dynamic context with variable user flows in terms of VNF type and location traversed and varying user types in the network. However, there is a gap in addressing the cases where the type of SC requested is variable with varying arrival rates of users and active user requests. We address this gap in the existing literature by proposing an algorithm which performs dynamic VNF placement considering the SC requested by the users while minimizing the number of VNF instances as well as the blocking probability. Moreover, the algorithm also considers bandwidth constraint of links, computational capacity of the NFV-nodes and latency requirements of the requested SC. We propose four versions of the algorithm with varying levels of SC-awareness and the presence or absence of Latency Island concept. We simulate the different versions of our algorithm in the context of a 4-stage aggregation metro network for remotely and locally served SC cases. We illustrate the results in terms of blocking probability and average number of activated VNFs.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis.pdf
non accessibile
Dimensione
1.33 MB
Formato
Adobe PDF
|
1.33 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/10589/135920