In recent years, cloud computing have emerged as a promising new execution environment for computational fluid dynamics (CFD) simulations. CFD field requires massive computational power and High-Performance Computing (HPC) solutions. Cluster and grid have been the answer for massive computing needs for a long time. Cloud, on the other hand, is eliminating the need for large upfront capital expenses, providers of IaaS clouds offer their customers on-demand instant access to unlimited computational resources. However, despite the growing popularity of cloud computing and its possible benefits, it still remains unclear whether Cloud can provide a viable alternative to cluster for HPC applications. This thesis analyse the efforts and challenges for porting and deploying CFD application to the cloud. In addition, we evaluate if IaaS clouds can provide sufficient capacities for running CFD applications, and compare its efficiency to traditional cluster. In particular, this thesis introduces TransAT Cloud version which enables running CFD simulations on remote dynamic resources in the context of parallel processing. Furthermore, in order to investigate challenges and bottlenecks that CFD simulations face in the cloud, this thesis introduces a new benchmark framework. Moreover, this thesis provides performance evaluation of IaaS clouds using data collected with this framework.
Negli ultimi anni, il cloud computing si è contraddistinto come nuovo ambiente di esecuzione per simulazioni di fluidodinamica computazionale (CFD), campo che richiede un’elevata potenza di calcolo ed altissime prestazioni (HPC). Cluster e Grid sono stati la risposta a queste esigenze di elaborazione per lungo tempo. Il Cloud, d’altra parte, sta eliminando la necessità di investire grandi capitali nelle fasi preliminari di un progetto, in quanto i fornitori di Cloud IaaS offrono ai loro clienti accesso istantaneo on-demand a risorse computazionali illimitate. Tuttavia, nonostante la crescente popolarità del cloud computing e dei suoi possibili benefici, rimane ancora poco chiaro se e quando il Cloud sarà in grado di fornire una valida alternativa al clustering per le applicazioni HPC. Questa tesi analizza le necessità ed i benefici di porting e distribuzione di applicazioni CFD al cloud. Inoltre, si valuta se il Cloud IaaS sia in grado di fornire risorse sufficienti all’esecuzione di applicazioni CFD, confrontando la sua efficienza al clustering tradizionale. In particolare, questa tesi introduce il TransAT Cloud, che consente l’esecuzione di simulazioniCFDtramite l’allocazione dinamica di risorse remote, sfruttando l’elaborazione parallela. Inoltre, al fine di approfondire le sfide ed i limiti imposti dalle simulazioni CFD, questa tesi introduce un nuovo riferimento di benchmark, oltre a presentare una valutazione delle prestazioni del Cloud IaaS utilizzando i dati raccolti.
Challenges of efficient Cloud utilization for fluid engineering simulations
KOLUNDZIJA, NELA
2013/2014
Abstract
In recent years, cloud computing have emerged as a promising new execution environment for computational fluid dynamics (CFD) simulations. CFD field requires massive computational power and High-Performance Computing (HPC) solutions. Cluster and grid have been the answer for massive computing needs for a long time. Cloud, on the other hand, is eliminating the need for large upfront capital expenses, providers of IaaS clouds offer their customers on-demand instant access to unlimited computational resources. However, despite the growing popularity of cloud computing and its possible benefits, it still remains unclear whether Cloud can provide a viable alternative to cluster for HPC applications. This thesis analyse the efforts and challenges for porting and deploying CFD application to the cloud. In addition, we evaluate if IaaS clouds can provide sufficient capacities for running CFD applications, and compare its efficiency to traditional cluster. In particular, this thesis introduces TransAT Cloud version which enables running CFD simulations on remote dynamic resources in the context of parallel processing. Furthermore, in order to investigate challenges and bottlenecks that CFD simulations face in the cloud, this thesis introduces a new benchmark framework. Moreover, this thesis provides performance evaluation of IaaS clouds using data collected with this framework.File | Dimensione | Formato | |
---|---|---|---|
tesi.pdf
non accessibile
Descrizione: thesisfinal
Dimensione
3.48 MB
Formato
Adobe PDF
|
3.48 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/10589/93442