Πανεπιστήμιο Μακεδονίας
Comprehensive formalization and modeling of the network and system virtualization process
Abstract
dc:descriptionThe virtualization concept has been of interest to the academic community and IT business sector for more than forty years, providing a different approach to the realization, administration and provision of physical resources. A paradigm is introduced by which the end user is not aware of the details of the underlying physical infrastructure. This approach has been warmly adopted in many fields of computing, including computer networks. The Internet – being the prime example of a large scale and complex network instantiation – has been the driving force behind the adoption of network virtualization as one of the key technologies in the field. In a parallel development, the convergence of communications and computing – two, previously, distinct worlds – has introduced the use of computing servers acting as active network elements (e.g., routers). The core networking support in these Virtual Network Environments (VNEs) is based on the IEEE 802.1Q VLAN implementation, where virtual network segments are established on top of physical switches – the latter being provided by the server’s hardware features. A thin software layer - the hypervisor - works as a virtual Ethernet switch, supporting queues for each VLAN in the system’s memory. Consequently, the network’s last-hop switch has been shifted from a dedicated active network element to become a characteristic of the hypervisor or of the physical server’s hardware. The aforementioned facts, along with the application of new technologies that lead to demand for novel services (wireless/mobile networking, Cloud computing, Networking As A Service, etc.) have considerably increased the complexity of VNEs. Cisco Systems projects that thirty seven billion intelligent devices (including smart fabrics and pills) will connect to the Internet by 2020 – dramatically increasing the traffic load and operational complexity of the “The Internet of Everything”. Efficient management of these architectures presumes the application of suitable information models. The latter provide the required standardization of abstraction of VNE elements and facilitate the construction and deployment of management methods. Despite the availability and benefits of several existing solutions, there are very few proposals that address the problem and the details of the introduction of computing servers in the network architecture, and of the hypervisor as a manageable entity, in particular. Moreover, existing solutions do not treat the hypervisor as a whole entity; rather they only, indirectly, reference its involvement via abstractions of hosted virtual machine operations. This results in impediments in managing modern VNEs; thus, in assuring the quality of the offered service on an end-to-end basis.In this dissertation we propose an information model that can conceptually abstract and describe physical or logical infrastructure elements participating in hypervisor host-based virtual network environments. In summary, the contribution of this dissertation is multifold: we initially provide a detailed review of existing modeling approaches employed in managing both VNEs as a whole, as well as autonomous system and networking components involved. We expose their pros and cons in a comparative review of the examined models and mechanisms using a conceptual categorization. We, then, describe our proposed solution: an information model that fits well within the context of a highly virtualized, host-based VNE, as those met in modern Cloud datacenters; then we apply a verification method for the validation of our approach. In particular, we draw upon one of the most important issues in Cloud environments – that of efficient resources allocation – and extend our model in order to include appropriate control logic and methods based on Statistical Process Control for managing hypervisor provisioned resources to a networking architecture. We test and validate our proposal extensively against actual VNE implementations. The verification provides a successful proof-of-concept of our proposed technique. Our proposed information model facilitates managing a VNE’s element, be that a networking, computing system or other hypervisor-provisioned resource. The management application does neither need to be aware of the specifics of the underlying virtualization platform, nor to understand it.
Degree
thesis:*- Grantor dc:publisher
- Πανεπιστήμιο Μακεδονίας
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kontoudis, Dimitrios
- Κοντούδης, Δημήτριος
Subjects
dc:subject × 26- Εικονικοποίηση
- Διαχείριση δικτύων
- Διαχείριση συστημάτων
- Επόπτες
- Πληροφοριακά μοντέλα
- Στατιστικός έλεγχος διεργασίας
- Νεφοϋπολογιστική
- Μηχανογραφικά κέντρα
- Δυναμική διάθεση πόρων
- Τραπεζική πληροφορική
- Virtualization
- Network management
- SYSTEMS MANAGEMENT
- Hypervisor
- Information modeling
- Statistical process control
- Cloud computing
- Datacenter
- Dynamic resources provisioning
- Banking information technology
- Φυσικές Επιστήμες
- Μαθηματικά
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Mathematics
- Computer and Information Sciences
Rights
dc:rights- Statement dc:rights
-
- BY_NC
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/38596