Smart bandwidth allocation for next generation networks adopting software-defined network approach

This data article contains information on a new intelligent bandwidth allocation model for future network (Smart Allocation). The included data describe the topology of the network testbed and the obtained results. Obtained data show the effectiveness of the proposed model in comparison with the MAM and RDM bandwidth allocation models. In relation to the performances evaluation, a variety of flows are used such as: voice over IP (VoIP), video, HTTP, and Internet Control Message Protocol (ICMP). The evaluation criteria are: VoIP latency and jitter, Peak Signal to Noise Ratio (PSNR) video, retransmission video, goodput, HTTP response page, and the Round-Trip Time (RTT) ICMP delay. The presented data are extracted based on simulation.


Subject area
Network Management, Communication Network Architecture, Network Layer, Network Design. More specific subject area Quality of Service, Software-Defined Network, Next Generation Networks. Value of the data Dynamic bandwidth allocation is one of the major concerns in the networking and telecommunications sector.

Type of data
The proposed model allows equitable distribution of bandwidth resources over different flows with different priorities.
The proposed model is tested for the next generation computer networks. It can be deployed in industrial networks.
These simulation data can be used as references for future work related to adaptive bandwidth management.
The proposed model is deployed at a controller; this allows researchers in the Software-Defined Network (SDN) axis to adopt it in order to optimize the performance of their networks.

Data
With the emergence of new flows (real-time, streaming, gaming, transactional, and file exchange), optimal bandwidth management has become one of the major concerns for both the telecommunication industry actors and the researchers. The Quality of Service (QoS) has become a necessity in the Next Generation Networks (NGN). Several algorithms have been proposed for dynamic bandwidth management while offering clients the solicited QoS level and guaranteeing operators the optimal use of their network infrastructures. The Maximum Allocation Model (MAM) [1] used to distribute in a fixed and absolute way the bandwidth between the different flows. Among the major limitations is the inefficient exploitation of the network resources, which means, even if no flow is present, its bandwidth cannot be temporarily re-allocated to others. The Russian Doll Model (RDM) [2] allows to correct this limit, allowing a bandwidth allocation of a flow for the benefit of another. Except that this allocation can be made just from a low priority flow to another high priority flow and not the reverse. This in fact can condemn the low priority traffic; it cannot be executed in the network. The AllocTC-Sharing [3] is a solution for allowing an allocation in both directions (from lower priority to higher priority and vice versa). However, the models mentioned above take into consideration the bandwidth consumption rate to make the adaptation. The Smart Allocation model makes it possible to take into consideration other criteria related to flows, for example latency, loss rate, and retransmission. The efficiency evaluation of a model must be carried out by increasing the load and varying the flow nature. The materials used in the experiment are the Cisco XR and Cisco 7200 IOS routers. The Smart Allocation model was deployed on a centralized controller, connected to routers through a 100 Megabit UTP link. The server contains a video sequence of 720 pixels' resolution. This video will be broadcast to the users to measure the quality of the video in the three bandwidth adaptation models.

Experimental design, materials, and methods
The experimental data shown in Table 1 includes the following metrics: VoIP latency: the delay between sending a packet and receiving it. VoIP jitter: the duration between the sending of two successive packets.    We cannot talk about the QoS without dealing with the video quality. The data in Fig. 2 illustrates the Peak Signal to Noise Ratio (PSNR) of a video sequence intercepted by the Smart Allocation, RDM, and MAM models. These values are obtained from the MSU Video Quality Measurement Tool. The evaluation methodology consists of comparing the quality of the video intercepted by the destination, based on a high-quality file stored in the server. The attached CSV file contains the numeric values used to generate this graph.
In order to qualitatively evaluate the data, Fig. 3 illustrates a microscopic capture of a video sequence number 87. The Fig. 3a illustrates the image quality with the Smart Allocation model, Fig. 3b with RDM, and Fig. 3c with MAM. The lines represent the glitches and distortions.
Other data was generated to compare the QoS adaptation models; we mention the goodput and the retransmission. The goodput means the amount of information received by the application layer of the OSI model without including the size of the lower layer headers. The higher the goodput data, the better the quality of the intercepted video. Fig. 4 illustrates the goodput data (in megabits per second) of the three models for the same video sequence used to measure the above criteria.
The packets retransmission is one of the factors that influence the transmissions quality in general and specifically the video quality. Fig. 5 illustrates the retransmissions number in the three models of the same video sequence.
The effective retransmission represents the number of packets successfully retransmitted.