Please use this identifier to cite or link to this item:
|Title:||Trust-aware Life-cycle Management of Federated Edge Clouds|
multi criteria decision making
|Abstract:||In the 5G and IoT era, service delivery and management models become more and more complex, by combining a wide variety of technologies and devices. The complexity of the new service delivery models raises new challenges for the fulfilment and management of the service life-cycle, from service discovery and selection, to orchestration, management and monitoring. In addition, in such context, where unknown users, devices and providers need to interact to achieve service fulfillment, it is critical to provide a sense of trust between users. The motivation for this dissertation is to tackle the new challenges arising in the service life-cycle management and enable trust in such scenarios. In the first part of this dissertation, our main focus is to assess services over heterogeneous resources, considering Quality of Service (QoS) and Quality of Experience (QoE) by combining monitoring data, the subjective perception of various users and their unique requirements. To achieve this, we aim to quantify the service assessment with the above criteria, as a reputation score value, that represents the performance and reliability of the service in a fair manner, while guaranteeing protection from malicious actors trying to manipulate it. This value can enable trust towards service providers while assisting users in the service selection process taking into account their unique priorities. To achieve the aforementioned goals, we developed collaborative Service Level Agreements (SLA) and Reputation Trust Management (RTM) frameworks with focus on cloud federations. Cloud Federation is the deployment, management and collaboration of several cloud computing services. It can integrate private, community, and public clouds providing multiple benefits such as scalability, fault tolerance and elasticity to the cloud environment while raising new challenges, namely interoperability issues of heterogeneous clouds and lack of trust among different providers. In this environment, we use SLA, in order to allow users to enforce guarantees about the performance of consumed services. We extend Key Performance Indicators (KPIs) assessed by the SLAs to surpass the limited availability KPI provided by most cloud and service providers at the moment, towards better addressing the specific requirements of each user. Furthermore, for the RTM aspect of our collaborative framework, we focus on Multi-Criteria Decision Making (MCDM) techniques for the quantification of trust and the computation of the reputation values. We adopt MCDM methodologies due to their capability of receiving as input multiple KPIs combining QoS and QoE metrics. In our research work, we modified and extended two known scalable fuzzy-based MCDM techniques, namely Fuzzy VIKOR and Fuzzy Analytic Hierarchical Process (AHP). The first one is horizontal while the latter one is a hierarchical MCDM system. The key contributions of our extensions in those systems mainly refer to their capability of facilitating inputs, not only limited to fuzzy numbers, but also to include linguistic and binary values. In this way, we can combine a wide variety of QoS and QoE aspects in the computation of the reputation score value, guaranteeing that a high reputation value represents wholesomely the provider's performance and reliability. We also modified those techniques to allow users to provide custom weights in the evaluation process to express different requirements. Last but not least, we introduce in both approaches credibility mechanisms that compare the user's evaluation with SLA and monitoring measurements in order to detect malicious actors and protect providers from reputation score manipulation attempts. The implementation and experimental validation and assessment of our novel solutions validated the effectiveness of our SLA techniques and the performance of our proposed RTM solutions. In particular, in comparison with other known frameworks in the literature, our proposed approaches highlighted the importance of mixing several numerical and fuzzy metrics, in contrast to adopting numeric only input, as well as the necessity and effectiveness of our introduced credibility mechanisms. On the second part of the thesis, we address challenges in the service life-cycle management. Our main focus is to facilitate the life-cycle management of services in a multi-domain federated Edge Cloud environment and enable tenants to either offer services for off-the-shelf leasing or lease a service that suits their needs. Federated edge clouds in the scope of this dissertation, refer to multi-administrative collaboration of multiple providers willing to allow cross-domain communication to allow off-the-self leasing of services developed by either providers or tenants through an Network Service Marketplace (NSM). In order to fulfill the above objectives, we shift from trust management frameworks to trustless systems to assist the life-cycle management of services in this context. We use blockchain as our trustless system, as it enables parties without trust between them, to interact without a central trusted authority, communicating in a decentralized fashion, maintaining the same functionality, while all parties can retain the certainty for the outcome of the transactions. We used blockchain technology to develop a fully functional NSM for off-the-self service leasing. We utilize blockchain's distributed ledger as a distributed database between different providers and edge clouds, and we leverage smart contracts to enable the platform and the transactions. Using smart contracts we developed all the necessary interactions and transactions required in an NSM facilitating all steps in a service's life-cycle such as service discovery, selection, leasing, billing and most importantly resource orchestration. We provide a multi-domain architecture aligned with the ETSI-NFV standards, which is highly scalable and requires minimum resource and management and development overheads for new Edge Cloud providers. Furthermore, in order to fulfill the consumption of a leased service in the NSM we developed a novel Cross-Service Communication (CSC) orchestrator over Network Function Virtualization (NFV) reference architecture. Our orchestrator, assisted with the minimum information required for the operation by our blockchain-based NSM, offers the necessary abstractions and operations for an automated and seamless CSC orchestration. The data stored in the blockchain are the minimum required and contain mostly tags guaranteeing both data efficiency and data privacy regarding the services offered. The implementation and evaluation of our NSM and orchestrator validated that our proposed solution is highly scalable, and the orchestration time overheads added from blockchain, API and orchestrator interactions are negligible. In addition, the Cross-Communication orchestration requires no extra resources, while other proposals in the literature require intermediary services to function. Finally, we summarize the conclusions of our thesis and offer interesting ideas for future development and enrichment of our work, along with significant synergies that can occur between the different solutions presented in this dissertation.|
|Appears in Collections:||Διδακτορικές Διατριβές - Ph.D. Theses|
Files in This Item:
|Papadakis_Konstantinos-PhD_Thesis.pdf||3.36 MB||Adobe PDF||View/Open|
Items in Artemis are protected by copyright, with all rights reserved, unless otherwise indicated.