In spite of all the major research work previously mentioned, one very important piece of the puzzle is still missing. “Multicast Services” Virtualization, according to the best of the author’s knowledge, has not received the appropriate attention it is entitled to, and little work has been done in this field. According to the literature in the previous section on network
virtualization, none of them mentioned multicasting in network virtualization. This section reviewed the work related to the multicast virtualization, as the focus of this thesis is virtualization of multicast services in WiMAX networks. All the research done in virtualization did not allow for the interchangeable services between network operators because their purpose was the isolation of services. However, as multicast service is the focus of this thesis, and many users may request similar content, isolation was not considered.
There exist two scenarios for multicast service virtualization as stated in reference [96]. In the first scenario, both the source and receiver are on the same virtual network. The content distribution tree is based on a specific Virtual Network (VN) and protocol independent multicast (PIM) instances. The second scenario is when the source and receiver are in different virtual networks. These enable different users to access multicast content delivered by a multicast stream in another virtual network. Figure 3.8 below shows a scenario where a multicast server with a Network Interface Card (NIC) 802.1q is set up. It also shows an end- to-end virtualization in which a complete isolation of resources, networks and users is considered. For example, if a user is subscribed to a service which has a source in its network or another one, the service provided by any other network can be received. This generates separate streams to be sent to different Virtual Local Area Networks (VLANs).
Figure 3.8. Multi-NIC multicast source environment [96].
Figure 3.9 below illustrates a source shared by multiple virtual networks, where a central policy enforcement point (Service Edge) is used to allow the communication between the source and the client by the replication of multicast content.
Figure 3.9. Replicated multicast source environment [96].
Since the bandwidth resource is rare in the field of networking, the work in reference [97] proposed a virtualization framework for resource provision that enables services which require high bandwidth to coexist on a single infrastructure. To solve the problem of bandwidth, they developed an algorithm that reduces the number of servers to satisfy all requests of those services. However the study ignored bandwidth efficiency use and advantages of multicast services. This is why, with the use of virtualization and advantages of multicasting service on top of the theory of heterogeneous networks, multicast traffic can be delivered through a single virtual network - thereby providing a solution to the bandwidth management problem.
The work in reference [98] investigated the problem of mapping network resources in the perspective of virtual multicast service-oriented network with the concern of delay and delay variation constraints. The work also proposed connectivity architecture for better network services for a cost effective method that supports the delivery of IPTV over wide area IP multicast that runs over a reliable virtual network. The connectivity layer was introduced between service layer and infrastructure layer which uses virtual link purchased from the infrastructure provider. The objective of connectivity architecture for better network service is to improve coverage, reliability and network performance. Investigation of mapping multicast service in virtual networks was done, taking into consideration the delay and delay variation [25]. The results have achieved minimum load balancing from service requests and thus increased the feasibility ratio of virtual multicast network requests. However their study did not focus on how to deliver the content to users [25]. This thesis considers the number of flows and virtual networks to allocate flows to a virtual network. However the mapping of multicast services does not show how to efficiently use the bandwidth in a virtualized
environment, which is the focus of this thesis. Table 3.2 summarises briefly the above stated network virtualization techniques.
Table 3.2. Comparison of network virtualization techniques
Approaches Proposed techniques Advantages Limitations Resource discovery and allocation in network virtualization with mainly heuristic solutions and rarely MILP [21] -Heuristic, MILP, Graph theory approaches for resource discovery and allocation.
-However, since the problem is NP hard, heuristics and meta- heuristics are and will remain useful for very large instances of the problem.
-Cooperation of multiple network operators in virtual environment was not considered. This takes much time for
resolution. An algorithm for virtual network embedding in multi-hop wireless network [22] -Graph theory. Heuristic solution.
-To optimise the revenue of the provider. -A single virtual network was considered hence cooperation of virtual network does not exist. Traffic classification based on QoS and allocation to virtual networks [23] -Machine learning techniques such as Naïve Bayes and Decision Tree for traffic classification and allocation.
-Effective classification process and traffic allocation. -No involvement of heterogeneity of the networks. Virtual network resource management [24] -Enables SPs to bind VNRs rented from heterogeneous NPs and facilitates NPs to perform cost-efficient allocation of VNRs. -Accelerate the realisation of virtual resource sharing in the future Internet business marketplaces.
-Uncertainty of virtual resource availability and lack of stringent network robustness for real-time applications. No clear policy of resource management. No efficient use of bandwidth.
Approaches Proposed techniques Advantages Limitations Switch virtualization that enables multiple network experiment to run on the same infrastructure [77]
-A set of flows forms a slice which is assigned minimum data rate by a FlowVisor. -Enable hardware forwarding plane to be shared among multiple logical networks, each with distinct
forwarding logic.
-No policy was put in place on how to assign the bandwidth to slices.
Online karnaugh -map embedding algorithm [80]
-An online algorithm was used to schedule Virtual Networks (VNs) requests using karnaugh-map to embed the Virtual network requests based on time window, buffer, and queuing.
-Efficient performance. -No cooperation between mobile network operators. Virtual network control using intelligent management and cognitive methods [82] -Dynamic creation of virtual networks based on the traffic flows, requirements of resources and the physical network.
-Offers improved virtual network capacity
-However, it does not show how bandwidth is efficiently used and does not discuss multicast flows Game theory and
Nash Equilibrium. The concept was used to manage resources [83], [14], [15]
-Using game theory to enable interaction between InPs and SPs. Using Nash Equilibrium to allocate bandwidth between virtual networks. -Efficiency bandwidth allocation. -Static allocation of traffic to virtual networks, while in this thesis virtual networks are selected dynamically. Spectrum sharing between virtual network operators using LTE [86] -Multiplexing gain based on network operators business policies like the current traffic load of the virtual network and the statistical characteristics of real-time traffic.
-Observe the statistical characteristics of real- time traffic for the accurate spectrum estimation required to each virtual operator.
-The virtual network selection and allocation of traffic to virtual networks which improves the bandwidth use efficiency was not considered.
Approaches Proposed techniques Advantages Limitations A contract based hypervisor algorithm estimates needed bandwidth of virtual operator [4] -The information about user channel conditions, loads, priorities, QoS requirements, contract policy of each of the virtual operators is used to schedule the air interface resources between virtual operators.
-Enhancement of the overall resource use. Possibility of opening the market to small operators.
-No allocation of traffic to virtual networks, no QoS guarantee policy was provided. Mapping multicast services to virtual network [25] -Mapping of multicast services was done based on delay and delay variation.
-It offers minimum load balance.
-Bandwidth is not efficiently utilised.
3.7. Summary
This chapter reviewed the related literature and studies the concept of virtualization in previous research. Most related studies analysed virtualization where network resources were shared, but did not show how the resources were shared among the virtual networks. The network selection of virtual network for efficient bandwidth use was ignored in the literature. In addition, the chapter discusses works on virtualization of multicast services. Very few scholars proposed how multicast service virtualization can be implemented, and the cooperation among network operators was ignored.
CHAPTER 4
4 WIMAX MULTICASTING
VIRTUALIZATION
4.1. Introduction
The previous chapters discussed the background and related works guided by the research objectives of this thesis. However, the aim of this chapter is to design a virtualized multicast service framework that enables efficient use of network resources that support interchange of service delivery. This is worked out between multiple networks on a shareable network infrastructure and the development of an algorithm that efficiently allocates multicast traffic to virtual networks. In this chapter, an optimum multicast rate and scheduling of multicast traffic is also designed. The Generalized Assignment Problem in [99] is applied in order to model the allocation of multicast traffic to virtual networks because the GAP allocates exactly one flow to only one virtual network to maximise the total throughput of the assigned flow. This is done without assigning multicast flows to any virtual network which exceed the total bandwidth greater than the virtual networks capacity. Thus, MILP [62] is used by allocating a flow to a single virtual network to solve the flow allocation problems. MILP assigns binary digit 1 whenever a flow is allocated to the virtual network and otherwise digit 0.
The chapter starts with the model of virtualization for multicast services by using GAP and analyses how the GAP problem was mapped to linear programing problem. After this, the optimal solution of multicast flows allocation to virtual networks is described. Finally, the multicast scheduling is presented.