A Visualization Based Analysis on Dynamic Bandwidth Allocation Algorithms for Optical Networks

This paper presents the bibliometric analysis of Dynamic Bandwidth Allocation (DBA) algorithms in the telecommunication field using CiteSpace and Web of Science (WoS) analysis for the period of decade (2009-2018). The publication count and citation counts in the last five years have significantly increased reflecting this as active research. The visualization results identified the most influential DBA algorithms centered on 1) EPON technology addressing DBAs meeting QoS requirement (68 counts), 2) DBAs for Elastic Optical Networks (58 counts) and (3 DBAs for long reach PON (54 counts). Assi CM and Kramer G are the leading contributors and Plos One and Journal of optical communication and networking are proffered Journals. The correlation between the research contribution and expenditures countries do is 0.945. Wherein, China in countries and Beijing university of Posts and Telecommunications among research institutes contribute the most in the DBA PON research across the world.


Introduction
Dynamic bandwidth allocation (DBA) is a method, which allows traffic bandwidth capacity to be distributed on demand as well as reasonably among multiple users of that bandwidth in a common shared communications medium. This is one of the techniques used in bandwidth management and is quite similar as statistical multiplexing in concept. In which the sharing of a link adjusts to the instantaneous traffic requirements of the linked access points in a certain way [1], [2] .
DBA algorithm uses many shared network characteristics: (1) Usually all users do not really remain connected to the network at one time (2) And even if connected, users do not send data (or voice or video) at all times (3) Most of the traffic volume is bursty in nature i.e. gaps exist between information packets which can be packed with many other data traffic. Different network protocols have completely different methods of implementing bandwidth allocation efficiently. These techniques are indeed typically generally defined in standards that are developed by regulatory bodies such as the ITU, IEEE, FSAN, or IETF. One excellent example is defined in the ITU G.983 specification for the passive optical network (PON) [3], [4].
DBA methods have emerged with mechanisms for sharing limited resources in a rapidly growing number of users in such access networks. These networks have also improved considerably over the years by increasing their bandwidth. Furthermore, QoS parameters are included in the bandwidth sharing procedure. These networks have however improved substantially over the decades by continuing to increase their bandwidth. Furthermore, QoS specifications are included in the bandwidth sharing operation [5], [6].
Since telecommunication, network & systems are one of the major research fields with incredibly fast-changing literature where almost every day new areas and terms continue to emerge. Keeping track of each gradual improvement in one's particular research subdomain can be very difficult. Existing detailed papers on the subject cover 100-200 maximum papers and present the analysis dependent on the one's experience, which can differ, from one expert to other. In contrast, bibliometric techniques use co citation analysis, which analysis the frequency of cocited papers in the papers to reveal its domain, present, past domain and its active or dead periods. Keeping in mind, this research contributes by presenting the bibliometric analysis of DBA algorithms in the telecommunication field. Research records for the period of decade (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) for this analysis are sought from Web of science. In the rest of the article, Section II presents the data, methodology, Section III presents findings in details, and Section IV contains conclusions.

Data and Methodology
The research data for this paper are retrieved from the Clarivate analytics -Web of Science (WoS) database 8 Jan2019 for the period from 2009 to 2018-A decade. The bibliographic data contains the complete record, cited references of articles, conference proceedings, review papers in English-language, summarized in Figure 01(a-b), and Table 01. The search was carried out on all four WoS indices, including SCI-EXPANDED, SSCI, A&HCI, ESCI and is fine-tuned which exclude data from completely irrelevant disciplines like Biology, Psychology etc. We obtained a record of total 2421 research publications on topic selection "Dynamic bandwidth allocation OR DBA" was downloaded from 2009 to 2018. The records contain information related to authors, title, years, and research institutions.  The record analysis per year confirms that there is growing research studies being done on the efficient distribution and utilization of bandwidth allocations. With the rapid development of information technology over the past several decades, the scientific visualization of bibliometric analysis has been achieved. The record analysis per year confirms that there is growing research studies being done on the efficient distribution and utilization of bandwidth allocations. With the rapid development of information technology over the past several decades, the scientific visualization of bibliometric analysis has been achieved. CiteSpace [7] has diverse visual analytic functions and facilitates the visualization of knowledge domains with clarity and interpretability of visualizations compared to the other existing visualization tools like VoS, Pajrek etc. [8]. CiteSpace is especially useful in identifying an intellectual base, emerging trends of topics, hotspots, and landmarks allied with various publications in a group of publications, and subsequently generating different visualization graphs or illustrations to represent the patterns of scientific literature in a specific domain such as [9], [10]. In this paper, we use CiteSpace and WoS analysis, focusing on detailed co-citation analysis, Author, Country and Research Institute category.

Results Generated from Document Co-Citation Analysis
The concept of CiteSpace is primarily based on general theory of co -citation analysis and pathfinder network scaling algorithms, which allow CiteSpace to pinpoint the major development pathways, and trends of a particular subject. Intellectual turning points actually play major roles in development and formulation of scientific domain. By recognizing such turning points, CiteSpace can find one subject's development path, which is useful for researchers to comprehend the topic and catch trending research issues or topics. Cited reference is a major aspect of co-citation visualization, which can properly categorize the most influential studies in a particular research domain. The time frame is set at 1 per slice from 2009 to 2018 for all the results findings of this paper and relative node type is changed in each analysis. Here we select node type as cited reference, and the strength among links measured by the Cosine metric: Where Cx Cy show the co-citation number between two papers x & y and denominator shows the times cited of the two papers respectively. The top 50 citations within each time slice were used for the analysis. Figure 1 shows the detailed outcome of ACA, i.e. co-citation network including 405 nodes and 1294 links. Modularity Q value (quality of cluster network formed) of the findings is equal to 0.726, which denotes that the network is reasonably divided into tightly coupled clusters. The mean silhouette value (indicates the similarity in the clusters) of 0.4108 indicates that the homogeneity of the clusters is fair.
Links in document co-citation networks convey the frequency of citing two articles together in another article in a data set [11]. Each dot in the visualization symbolizes a node in the network that is a cited reference. The merged network with nodes and links shows the development of a knowledge domain over a specific time by highlighting significant publications with labels. These significant publications, given Table no 02, are the highly cited references that can be considered as landmark papers in the knowledge domain.
In Figure 2, citations with large nodes represent the frequently cited publications and suggest that these papers contribute substantially to the DBA research. Therefore, Assi [12] is one of the frequently cited publications important for constructing the DBA research base. This article is published in IEEE journal on selected areas in communications and present differentiated services supported DBA algorithm that distributes bandwidths effectively and reasonably among end users. This DBA proved to exceptional research contribution with acceptable local queue management to mitigate inappropriate behaviour for certain traffic classes. Table. 2 mentions the titles, citation count (CC) and reference of the landmark papers in the field of DBA for the aspirant researchers.

Identification and Interpretation of Clusters
Cite Space analysis facilitates more precise ways to identify some prominent groups in a data set, known as clusters. Each cluster distinguishes a different domain [12]. Modularity and the mean silhouette scores indicate the properties of each cluster. If the modularity is relatively "Distance-adaptive spectrum resource allocation in spectrum-sliced elastic optical path network" [17] high, the network is divided into loosely coupled clusters. Further, a higher silhouette score suggests that the homogeneity of the cluster is high [9]. Cluster labels are assigned from each cluster's noun, and the noun phrases are retrieved mostly from the publications ' titles, keywords, and abstracts. Top-ranked noun phrases were chosen as the labels for the cluster. Specialized metric, namely, Log Likelihood Ratio (LLR) is used to identify the most significant clusters of DBA mechanisms and their most significant terms. LLR test identifies the uniqueness of a term to a specific cluster [7]. Cluster identification is shown in figure 03 and Analysis of cluster information, given in Table 3

Citation Bursts
Citation bursts indicate a specific duration in which an abrupt change of the frequency takes place in the research area. The results are extracted from above ACA technique where the red circle around the node represents the significant citation burst, indicating that citations of this node have emerged rapidly in a particular time period. Results in Table 4

Journal and Conferences Co-Citation Analysis
Here, we present journal and conferences co-citation analysis for the identification of interrelated core journals and conferences in the literature of "Dynamic bandwidth allocation OR DBA". Based on record of 2421 research publications retrieved from WOS for the period of a decade (2009-2018), we retrieve a list of top five journals and conferences in the literature, given in

Country wise Research Contribution
We used here WoS analysis to find out the leading countries as per their research contribution on this selected literature topic of DBA. Further, Gross expenditures on research development (GERD) [22] is also important aspect to analysis the research contribution from each country. GERD from China and Canada are less but research counts are higher than USA and Japan respectively. From Table 6 of top five countries, we come to know that China tops with research contributions and whereas South Korea is the top fifth country.

Conclusion
In this work, we collected 2421 records on "Dynamic bandwidth allocation OR DBA" algorithms from WoS, and conducted a visualization analysis on these studies using one mainstream co-citation analysis tool-CiteSpace and WoS. From the visualization results, we got some enlightening observations in terms of influential DBA studies in EPON, EON and Long reach PON [12]- [15]. The burst analysis revealed Assi CM and Kramer are the leading and highly cited DBA researchers. Plos one, JOCN and JLT journals are the preferred choice for the journal publication and IEEE ICCC for conference publication. China leads with 24 % of the research contribution made among the countries and in the number of research institutes. In the list of leading research institutes on the PON, BUPT leads the rest with 89 highly cited research counts. Above Co-citation, analysis and WoS Analytics would support interested researchers to make further studies on DBA algorithms.