In recent years, virtualization has changed the way network operators provide services to their customers. In particular, Network function virtualization (NFV) has attracted considerable attention as it provides a major shift in the provisioning of telecommunication services by decoupling network functions from dedicated hardware devices, which results in operational expenditure (OpEx) and capital expenditure (CapEx) reduction and increase in service agility. NFV relies on Virtualized Network Functions (VNFs) and, by placing them on the network nodes and chaining these VNFs together in a specific order, Service Chains (SCs) are formed each providing a specific service. When deploying a service chain, it is important to consider the case in which a failure will happen in one of the network components. In other words, to guarantee service continuity, the mapping of VNFs to network nodes equipped with storage and computational resources (referred to as “NFV-nodes”) and the routing of traffic across these VNFs must be done in a way to guarantee a backup SC configuration in case of failure. Different protection approaches can be followed to protect the SC against link and/or node failure. Previous works in the literature have addressed the problem of protection for SCs by proposing different Integer Linear Programming (ILP) models, ensuring resiliency against single-link and/or single-node failures. The models try also to meet SCs latency requirement and activate the fewest number of NFV-nodes. Previous studies also proposed algorithms to minimize the physical resources consumption while guaranteeing the required high reliability. However, no previous study has investigated protection strategies for dynamic VNF placement and service chaining considering dynamic traffic conditions. We address this gap in the existing literature, and we propose three different heuristic algorithms which provide three different methods of protection for dynamic VNF placement. Hence, providing protection for SC requests generated dynamically by the users in the network, consider bandwidth constraint of links and computational capacity of the NFV-nodes. We demonstrate the performance of our algorithm considering different metrics as blocking probability, total number of NFV-nodes with at least one VNF instance placed on them and latency violation. Our results show that using end-to-end protection strategy, we consume more resources in the network which results in having higher blocking probability and higher activated NFV-nodes. while virtual-link protection requires less resources and has the lowest value in terms of blocking probability, activated NFV-nodes and latency violation, and the activated NFV-nodes and blocking probability values of virtual-node protection lies between the values of end-to-end and virtual-link protection.
Negli anni recenti, la virtualizzazione ha cambiato il modo in cui gli operatori di rete forniscono servizi ai loro clienti. In particolare, il Network Function Virtualization (NFV) ha attratto un’attenzione particolare, in quanto ha provocato un profondo cambiamento nella fornitura di servizi di telecomunicazioni, disaccoppiando le funzioni di rete e risultando sia in una riduzione di operational expenditure (OpEx) e capital expenditure (CapEX), che in una maggiore agilità di servizio. NFV si basa su Virtualized Network Functions (VNFs) che, venendo posizionate all’interno dei nodi di rete in uno specifico ordine, formano le Service Chains (SCs). NFV permette di consolidare diverse funzioni di rete sullo stesso server fisico, è tuttavia importante considerare il caso in cui un malfunzionamento avvenga inn uno dei componenti della rete. In altre parole, per garantire la continuità del servizio è necessario mappare le VNF nei nodi di rete in maniera tale da assicurare la continuazione del servizio in caso di malfunzionamento. Esistono diversi approcci per proteggere le SC contro malfunzionamenti di collegamento o di nodo. I precedenti lavori in letteratura hanno affrontato il problema proponendo modelli di Integer Linear Programming (ILP) che assicurano resilienza in caso di un singolo malfunzionamento di link o nodo. Questi modelli tengono anche conto della latenza delle SC e cercano di attivare meno nodi NFV possibili. Ulteriori ricerche propongono algoritmi per minimizzare il consumo di risorse fisiche, soddisfacendo l’affidabilità richiesta. Tuttavia, c’è una mancanza di studi su strategie di protezione in caso le condizioni variabili della rete siano tenute in conto. Il nostro lavoro propone quindi tre differenti algoritmi euristici basati su tre diversi metodi di protezione per VNF in ambienti dinamici. In particolare, la protezione è dedicata alle SC richieste in maniera dagli utenti della rete e considerando sia i limiti di banda dei collegamenti, che la capacità computazionale dei nodi VNF. Analizziamo infine le prestazioni dei nostri algoritmi considerando diverse metriche: probabilità di blocco, numero complessivo di nodi su cui sia attiva almeno un’istanza di VNF e violazione dei requisiti di latenza. I nostri risultati dimostrano che usare strategie di protezione end-to-end consuma più risorse di rete, con un conseguente aumento della probabilità di blocco e maggiori nodi NFV attivati. Allo stesso tempo, le tecniche di virtual-link protection richiedono minori risorse e hanno i valori più bassi di probabilità di blocco, numero di nodi NFV attivati e violazione di requisiti di latenza. Infine la probabilità di blocco di tecniche di protezione virtual-node si posizione tra i valori corrispondenti a protezione end-to-end e virtual-link.
Protection strategies for dynamic VNF placement and service chaining
TAMIZI, MOHAMMADHASSAN
2018/2019
Abstract
In recent years, virtualization has changed the way network operators provide services to their customers. In particular, Network function virtualization (NFV) has attracted considerable attention as it provides a major shift in the provisioning of telecommunication services by decoupling network functions from dedicated hardware devices, which results in operational expenditure (OpEx) and capital expenditure (CapEx) reduction and increase in service agility. NFV relies on Virtualized Network Functions (VNFs) and, by placing them on the network nodes and chaining these VNFs together in a specific order, Service Chains (SCs) are formed each providing a specific service. When deploying a service chain, it is important to consider the case in which a failure will happen in one of the network components. In other words, to guarantee service continuity, the mapping of VNFs to network nodes equipped with storage and computational resources (referred to as “NFV-nodes”) and the routing of traffic across these VNFs must be done in a way to guarantee a backup SC configuration in case of failure. Different protection approaches can be followed to protect the SC against link and/or node failure. Previous works in the literature have addressed the problem of protection for SCs by proposing different Integer Linear Programming (ILP) models, ensuring resiliency against single-link and/or single-node failures. The models try also to meet SCs latency requirement and activate the fewest number of NFV-nodes. Previous studies also proposed algorithms to minimize the physical resources consumption while guaranteeing the required high reliability. However, no previous study has investigated protection strategies for dynamic VNF placement and service chaining considering dynamic traffic conditions. We address this gap in the existing literature, and we propose three different heuristic algorithms which provide three different methods of protection for dynamic VNF placement. Hence, providing protection for SC requests generated dynamically by the users in the network, consider bandwidth constraint of links and computational capacity of the NFV-nodes. We demonstrate the performance of our algorithm considering different metrics as blocking probability, total number of NFV-nodes with at least one VNF instance placed on them and latency violation. Our results show that using end-to-end protection strategy, we consume more resources in the network which results in having higher blocking probability and higher activated NFV-nodes. while virtual-link protection requires less resources and has the lowest value in terms of blocking probability, activated NFV-nodes and latency violation, and the activated NFV-nodes and blocking probability values of virtual-node protection lies between the values of end-to-end and virtual-link protection.File | Dimensione | Formato | |
---|---|---|---|
Mohammadhassan_Tamizi.pdf
non accessibile
Dimensione
4.93 MB
Formato
Adobe PDF
|
4.93 MB | Adobe PDF | Visualizza/Apri |
Mohammadhassan_Tamizi.pdf
non accessibile
Descrizione: Second Version
Dimensione
6.92 MB
Formato
Adobe PDF
|
6.92 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/10589/154159