Naslov
Scalable Monitoring for Energy Efficient Cloud Environments

Autori
Mastelić, Toni

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, doktorska disertacija

Fakultet
Fakultet Informatike

Mjesto
Beč, Austrija

Datum
22.10

Godina
2015

Stranica
197

Mentor
Brandić, Ivona

Ključne riječi
cloud computing ; monitoring ; scalable ; energy efficiency

Sažetak
The introduction of the Internet in the early 1970s represented a major shift in the way people communicate. It provided a global communication network that allowed data to be exchanged between network nodes which performed computation on the data and stored it. However, with the growth of the Internet, requirements for more computational power and storage capacity grew as well. This made the local infrastructures insufficient, hence leading to the introduction of Cloud Computing. Today, the Cloud represents a well-established paradigm for providing computational and storage resources to remote users in form of online services. However, due to its seemingly infinite capacity and popularity growth, the Cloud´s underlying infrastructure reaches ultra-scale size, which makes Cloud services extremely challenging to manage. Moreover, along with service provisioning, Cloud management deals with the performance and energy efficiency trade-offs due to its variable workload and ever-growing energy consumption. The main instrument for tackling these trade-offs is management itself. It includes monitoring of the underlying infrastructure in order to obtain intelligence and utilizing it to improve energy efficiency. Monitoring however implies collecting enormous amounts of data from an ultrascale environment, which makes monitoring system itself a major energy consumer and a challenge to operate. Additionally, the volume of the collected monitoring data does not necessarily translate to useful information. The cause of this is data redundancy, noise within the data itself, and lack of adequate metrics for measuring environment values. In this thesis, we propose a novel energy efficient monitoring system that scales with a Cloud environment and provides instruments for filtering and analysing monitoring data on the spot. We also introduce novel metrics used for analysing benefits and trade-offs of Cloud environments, including the performance guarantees expected by a user. Finally, we demonstrate the usage of the monitoring system and metrics within the Cloud management system and evaluate our contributions by comparing them to the existing practices and research breakthroughs.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA