Zur Kurzanzeige

Automated Provisioning of Fairly Priced Resources

dc.contributor.advisorFu, Xiaoming Prof. Dr.
dc.contributor.authorSridhara Rao Prasad, Abhinandan
dc.date.accessioned2018-07-03T09:53:41Z
dc.date.available2018-07-03T09:53:41Z
dc.date.issued2018-07-03
dc.identifier.urihttp://hdl.handle.net/11858/00-1735-0000-002E-E43C-0
dc.identifier.urihttp://dx.doi.org/10.53846/goediss-6955
dc.language.isoengde
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subject.ddc510de
dc.titleAutomated Provisioning of Fairly Priced Resourcesde
dc.typedoctoralThesisde
dc.contributor.refereeFu, Xiaoming Prof. Dr.
dc.date.examination2018-06-21
dc.description.abstractengIn recent times, cloud service providers are increasingly offering more complex services. These complex services are heterogeneous and composed dynamically from traditional services such as IaaS, SaaS, and PaaS to handle ad-hoc demands. Consequently, the current cloud is a complicated marketplace. Furthermore, resource prices influence the user demands, and user demands eventually drive resource provisioning. Correspondingly, resource pricing and provisioning are indispensable to each other. Hence, cloud service providers face the challenge of optimizing resource prices and provisioning. This challenge has attracted both industry and academia. However, most of the pricing approaches proposed and practiced achieve either efficiency or fairness. Thus, current pricing schemes reward either the service provider or the user. In resource provisioning, both industry and academia focus on addressing the issue of when to provision, while disregarding what to provision. Consequently, services are deployed using a single VM size for all components resulting in performance degradation and eventually leading to SLO violations. This dissertation proposes ARPP for the pricing and provisioning of cloud and edge resources to address the above-mentioned challenges. The ARPP pricing supports three pricing approaches dubbed RAERA, ERM, and OFM. RAERA is a robust optimization-based sealed auction proposed to address price uncertainty. ERM computes differential prices for buyers with SPLC utilities. OFM is an online Fisher market for pricing varying resources for varying buyers. Both ERM and OFM maximize NSW – a Pareto outcome between efficiency and fairness. The evaluation demonstrates the effectiveness and scalability of proposed approaches. In this dissertation, we propose RConf and RConfPD as an answer to automate the issue of what to provision. RConf finds an optimal configuration for maximizing the overall resource utilization of a complex service. Conversely, RConfPD trades off resource utilization for performance. Hence, it is appropriate for services with nearly instant provisioning requirements. Both the approaches estimate performance cost for arbitrary arrivals and departures using a robust queueing theory-based model. The experimental evaluations show the overall resource utilization improvement of 16-50% over one-size-fits-all solutions, and simultaneously deploys 22% of fewer resources. The ARPP proposed in this dissertation can be integrated with edge or cloud orchestrators. Moreover, the ideas presented can also be applied to (i) pricing and provisioning NFV service chains, (ii) building a single marketplace for cloud, edge, and fog resources and applying ARPPERM to price resources differentially according to resource types, and (iii) automatically selecting complex service type (cloud, edge or fog) and provisioning depending on user preferences.de
dc.contributor.coRefereeAschenbruck, Nils Prof. Dr.
dc.subject.engCloud computingde
dc.subject.engResource provisioningde
dc.subject.engResource pricingde
dc.subject.engFisher Marketde
dc.subject.engOnline pricingde
dc.subject.engConfigurationsde
dc.subject.engRobust optimizationde
dc.subject.engNash Social Welfarede
dc.identifier.urnurn:nbn:de:gbv:7-11858/00-1735-0000-002E-E43C-0-9
dc.affiliation.instituteFakultät für Mathematik und Informatikde
dc.subject.gokfullInformatik (PPN619939052)de
dc.identifier.ppn1025630556


Dateien

Thumbnail

Das Dokument erscheint in:

Zur Kurzanzeige