QoS-Aware Multicast for Crowdsourced 360° Live Streaming in SDN Aided NG-EPON

The IEEE 802.3ca Task Force group has recently approved the standardization of next-generation Ethernet passive optical network (NG-EPON) to satisfy future bandwidth demands. NG-EPON architecture classifies the optical network unit (ONU) transmission to allows simultaneous Wavelength-Agile (WA-PON) transmission which increases the network capacity (50 Gb/s) and flexibility. The recently introduced 5G services and related functional standards, new capabilities and Quality of Service (QoS) are imposed in the optical transport network. The Peer-to-peer (P2P) based live streaming application is the most popular multimedia application in access networks i.e., Facebook Live, YouTube Live, Twitch and other streaming, as well as video conferencing services are increasing especially in situations of global epidemics, so which requires a higher QoS. Therefore, media-based applications require adequate infrastructure resources to meet customer demands such as high bandwidth, low latency, high quality audiovisual, live streaming and services. Therefore, in this paper, we propose an SDN enhanced P2P virtual reality (VR) 360° live video streaming application NG-EPON multicast architecture based on 5G-enabled tactile internet (TI) and aims to improve the QoS, QoE and provide flexible services to the end-users. Finally, we proposed an SDN based TI-dynamic wavelength and bandwidth allocation (SD-TI-DWBA) mechanism. Simulation results show that our proposed architecture can improve the system performance and QoS metrics in terms of packet delay, jitter, system throughput, and packet loss.


I. INTRODUCTION
Great work is now underway to streamline fifth-generation technologies (5G) in supporting high-quality streaming media and entertainment applications. The significant amount of bandwidth demand and reduce delays due to increasing new technologies and services over the past few years, ultra-high definition video streaming(4K), virtual/augmented reality (VR/AR), mobile fronthaul, requires continuous evolution of the wireless network, and optical access will play a vital role in the different types of new services [1]. According to the Cisco Visual Networking Index by 2021, over 82% of global data traffic will be video content, growing from 73% in 2016. In addition, Worldwide, 0.9% of all entertainment traffic (Internet video and gaming) will be due to VR in 2021, up from 0.1% in 2016 [2]. Despite the above-mentioned statistics, the COVID-19 epidemic has led to an increase in the use of online conferencing for work-from-home and education purposes. In these pandemic situations, people are mostly using online meeting and online learning events platforms, such as google meet, zoom, Microsoft teams, Cisco WebEx etc. Currently, Internet peer-to-peer (P2P) usage is on the rise, flooding the internet backbone and stressing Internet Service Providers (ISPs). The live streaming applications have also been exploited to offering video services over the IP, i.e., such as PPLive, YouTube, Facebook, Instagram, Google meet [3]. However, a number of challenges have been identified and addressed in the design of P2P live video streaming systems such as high bit rate, end-to-end low latency, packet loss, network congestion, stream synchronization and service guarantee. The P2P live streaming service needs proper solutions to solve resource limitations in the access network [4]. P2P live streaming is a graceful service because it allows users to watch high-quality videos on the online and turn anyone into a content provider so they can share content with other users. Moreover, Broadband wireless technologies (such as 4G and 5G) enable users to send and receive multimedia traffic via IP-based networks while maintaining QoS and QoE, mobility, security, and interactivity.
5G enabled Tactile Internet (TI) imagines real-time monitoring, management and control of remote infrastructure. TI is expected to enable a plethora of new opportunities and applications that will transform our lives and economy [5]. Further, TI will provide a true paradigm shift from contentdelivery to skill-set delivery networks, thereby causing a revolution in every section of society [6]. The IEEE 1918.1 tactile internet baseline standard has provided the following various emerging applications and key metrics such as [7]: immersive VR, teleoperations, automotive, internet drones, interpersonal communications and live haptic-enabled broadcast. Currently, VR applications are widely used as entertainment forms such as gaming, 360° video streaming. Moreover, VR is used for military, education, healthcare(telesurgery), education, sports and many more industries are using [8]. The VR market expects revenue of 108 billion USD until 2021 [9] and VR HMD headsets are expected to increase fivefold, from 18 million in 2016 to nearly 100 million by 2021, the VR headsets are mostly connected to the smartphone and PCs., and the second trend is the availability of omnidirectional cameras, which make it easy to create customized 360° videos such as GoPro OmniALL and Insta 360° One [10]. As VR-based 360 • panorama video technology becomes more common, the need for efficient streaming methods for such videos arises which is provides the user with an interactive experience never imagined before [11]. There has been a significant increase in the production and use of omnidirectional or 360° videos, supported by the latest technological advances in networking and computing and by increasing users' interest in enriching their experience. Currently, many commercial broadcasters and video sharing sites are showing significant attraction in this sector and video sharing sites show considerable interest in this domain and the vital video platforms, such as YouTube, Facebook, Vimeo, Twitch, and ARTE, have put in many efforts into promoting 360° video services [12].
Passive optical networks (PONs) are becoming an important and reliable solution for various applications of access networks due to their high capacity and coverage capability. The different kind of PON technologies are there but an Ethernet passive optical network (EPON) with 1 Gb/s is widely accepted and dominant optical access technology for several years [13]. The cisco VNI is predicted the demand for high-speed data services is continuously increasing, and it is expected that global IP data traffic will attain 3.3 ZB by 2021 [14]. Therefore, the 1 Gb/s-EPON cannot satisfy the requirement of future applications such as mobile fronthaul, control office re-architecture datacenter (CORD), business users and network virtualization. To meet the future bandwidth demands, standardization has been recently approved by the 25 Gb/s and 50 Gb/s per wavelength IEEE 8023.3ca task force [NG-EPON], and newly introduces on ONU channel bonding [15,16]. Moreover, hybrid EPON has been classified into various design architectures depending on how bandwidth is managed over multiple wavelengths: 1) single scheduling domain (SSD), multi scheduling domain (MSD), and wavelength agile (WA). The dynamic wavelength and bandwidth allocation (DWBA) are a method that had proposed an NG-EPON architecture to satisfy the growing traffic demands in the future. The NG-EPON based recent DWBA schemes (First-Bit DBA and Water filling DBA are studied while considering one wavelength in the system [13,17,18]. Ref [19] introduce a new SDN based wavelengthagile WA-NG-EPON CORD architecture and water filling DWBA method that meets the QoS requirements of the tactile and non-tactile application on the cloud distribution nextgeneration optical access networks. The emerging Software Defined Network (SDN) is providing operators with improved resource provisioning, flexibility, and lower operational expenditures. SDN focuses on separating the control and data plane functions of the network, where the control plane determines the packet flow through the network, and maintains, controls, and programs the data plane. Furthermore, SDN aims to use the centralized programmable network model used by the OpenFlow protocol to modify the SDN mechanism in the network. The author [20] has proposed the SDN-EPON architecture and implemented the hard-coded dynamic bandwidth allocation (DBA) module in PON OLTs is modified with the advanced SD-DBA algorithm to accommodate additional global network information and higher granular traffic conditions in ONUs. The SDN based SIEPON architecture proposed and the SDN controller manages the resource utilization, flow monitoring, bandwidth assignment and QoS guarantees of the optical access networks [21]. Further, SDN and 5G based multimedia services are studied [22,23]. EPON has been extensively studied to tackle DBA and multimedia service architectural issues in [24][25][26][27][28]. However, to the best of our knowledge, few or no studies have been performed presently concerning SDN based P2P VR based live video streaming in EPON. Table 1 in some of the above works, illustrate the contributions of our current paper. The main contribution of this paper is as follows: DWBA scheme to support the QoS services in SD-NG-EPON. In the Proposed DWBA mechanism, we use a multi-wavelength sharing wavelength-agile mechanism are introduced. The proposed SD-TI-DWBA algorithm aims to improve the system overall QoS alleviating inter ISP-traffic. 5. Finally, we perform simulation experiments: the proposed SD-TI-DBWA scheme is compared to other non-TI-DWBA programs. The proposed SD-TI-DBWA scheme has been improved in terms of overall SD-NG-EPON multicast system delay and packet loss rate. The remainder of this paper is organized in the following manner. The 5G enabled VR-video streaming TI architecture is presented in section II. The proposed multicast video streaming system model is described in section III. Furthermore, the proposed SD-TI-DWBA model is described in section IV. In section, V discusses the performance evaluation and simulation. Finally, section VI brings our conclusion and future work.

II. VR VIDEO STREAMING TACTILE INTERNET ARCHITECTURE
The section aims to provide the overview of the VR based TI architecture and operations. TI applications are classified into two levels, such as 1) nature of the control environments (like physical-e.g., Telesurgery and Virtual-e.g., VR Gaming); 2) type of operator-teleoperator combination and the operator and teleoperator can be either human or machine. Figure 1 Illustrates the VR based 5G enabled TI low latency architecture. This architecture consists of three domains, such as 1) Tactile edge A master domain (controller); 2) Network Domain (communication) and 3) Tactile edge B of slave domains (controlled). These three domain detailed descriptions are explained give below.

A. TACTILE EDGE A MASTER DOMAIN
This primary domain is a human or machine (operator) and a human are a controller, a haptic device, often called the Human System Interface (HSI). In this architecture, HSI included a VR head-mounted display (HMD) device to streamline real-time video feed from the slave. The HSI is converted the human input signal to the haptic input signal through the various tactile coding techniques and also provides the haptic feedback from the slave domain to human. The haptic sensors and actuators are more important components of both master and slave domain communication systems and this device are used to send and receive the real-time physical sensation data (touch and feel). This is categorized by active and passive. Moreover, the multimodal sensor is used to improve the perceptual performance of audio and visual and haptic feedback.  This domain is used to connect the mater and slave domain via bilateral communications link over any wired/wireless/optical networking technology. TI require ultra-reliable and ultraresponsive network connectivity to transmit the real-time physical sensation data. The 5G networks must meet these requirements in order to support emerging TI applications. It provides less than 1 ms and 99.999% reliability and ensures proper operation of TI Edge/cloud computing reduces the latency and SDN provide the network more agile and flexible. The TI-support engine which provides the artificial intelligence and a knowledge base, and media, medical data analysis plays a vital role in stabilizing the system [29].

1) APPLICATION LOGIC
The master domain has a "controller" function with interfacing devices, multiplexing multimedia signals and interacting with restricted tactile devices.

2) MEDIA CODECS
It's provides digitalization of encoding/ decoding functions for haptic (kinesthetic and tactile), audio and video signals [30].

3) LEARNING ALGORITHMS
It monitors and modifies application parameters to meet the Quality of Experience Quality (QoE), regardless of any network delays and losses.

C. TACTILE EDGE B SLAVE DOMAIN
The slave domain consists of a teleoperator, such as a slave robot, and is directly controlled by the master domain through various command signals that define position and velocity. It means to record and stream the visual feed (usually a set of cameras that provide a full 360° view). The slave domain simultaneously receives VR data and transmits the force signals in the remote environment through kinesthetic force feedback data and the operator is fully immersed in the remote environment. Finally, realizing 5G enabled TI based VR 360° live streaming involves the following functional considerations: latency; reliability; availability; bandwidth; and scalability.

III. PROPOSED ° LIVE STREAMING SYSTEM ARCHITECTURE AND OPERATIONS
The proposed architecture supports low-latency-sensitive applications like tactile internet (P2P VR based 360° live video streaming, remote robotic surgery and more services) and mission-critical IoT services. Moreover, the architecture preserves the PON's unified design and takes advantage of the programmable state-of-the-art wavelength-agile (WA) PON technology, complemented by an effective resource provisioning system to accommodate all forms of media streaming services. Therefore, this MEC and software defined-based multimedia application architecture SD-OLT and SD-ONU module and tactile internet will increase the initial cost of NG-EPON development, although such modules may benefit ISPs, i.e. reducing and managing inter-ISP costs, and improving QoS and QoE for customers. This paper considered the in terms of following QoE, components such as synchronization, bitrate, signalling and propagation delays and quality level of consecutive chunks. point-to-multipoint (P2MP) topology, a logical topology emulation (LTE) feature is used that, depending on its configuration, can emulate either a shared medium or a pointto-multipoint (PtM) medium by using logical connection identifiers-LLID assigned to SD-ONUs. Moreover, note that IEEE 802.3ca next generation 25G-EPON has newly initialized an ONU Group link identifier (GLID), in this paper, we are using GLID (I.e., To aid in traffic management, the Nx25G-EPON framework supports grouping multiple LLIDs using the group connection ID (GLID)) [32]. A multi-point control protocol (MPCP) transmission message is used to transmit to the data, which is auto-discovery assigns the ONU LLID. Each ONU must register with the OLT within a specified time period, and the OLT will send the REGISTER message to configure the ONU LLID for each ONU. When the OLT has received the REGISTER ACK message, it is aware of both the ONU's MAC address and its LLID. Each ONU can have multiple LLIDs or GLID. Therefore, to multicast the live streaming channels, the channel GLID (CGLID) is defined; it is a unique GLID that is assigned to each requested channel. When the SD-ONU requests or leaves a channel, the SD-OLT assigns or deletes the SD-GLLID. ONU's. To control GLLID operations, a table is created in the reconciliation sublayer (RS) of each SD-ONU and SD-OLT (RS). SD-ONU contains the channel name, channel GLID, user MAC address, and user IP address fields, while SD-OLT contains the channel GLID, channel name, and SD-ONU GLID fields. Finally, the SD-OLT dedicates wavelength channels for tactile video streaming traffic and allows for the design and implementation of a bandwidth allocation scheme that enables wavelength agile (WA) [17]. This would improve the network performance and reduce the possible loss of bandwidth by using different wavelengths.

B. MULTICAST SDN CONTROLLER
The Optical infrastructure is governed by a centralized SDN controller that interacts with the agents located at the OLT and ONU. The Secure OpenFlow protocol is used to communicate between the SD-controller and the OF switches. The controller is capable of collecting network conditions and configuring the flow control tables of the switches. It is worth mentioning that, an EPON connection is used to establish the secure connection from the SD-controller to all ONUs. The flow tables are used by the OF-switch to forward packets. The flow table contains a flow entries list with match fields, a counter, and instructions and the arriving packets are compared to the match fields in each entry. Finally, the SD-controller is responsible for the DWBA, PON system activation and deactivation and integrated QoS management with the metro or core network. Therefore, SDN and NG-EPON optical access networks to conceive controllable video streaming systems capable of adapting to the dynamically fluctuating network conditions.
As is shown in Figure 3, the SDN-aided NG-EPON multicast system architecture is composed of a Media streaming server, SDN, Switches, SDN controller, SD-OLT, SD-ONU and End user devices. Media streaming server is responsible for video content storage and SVC encoding as well as streaming. The switches is responsible for communication between the SD-controller, media server and SD-OLT. The SD-controller, we customize the SDN controller, DWBA, and QoS routing for the SDN controller to assist us in controlling the network and selecting video streams. his module deploys the logical multicast trees generated by DWBA and QoS routing module into the OpenFlow network. Since the SDN controller, OpenFlow forwarding rules and configurations have multicast trees incorporated, the OpenFlow switches can collaborate to stream the video layers in the manner of multicast. When joining or leaving event occurs (I.e., User join or leave, group etc.) the multicast management module in SD-OLT changes the associated multicast trees and notifies SDN controller to reconfigure OpenFlow switches to forward packets according to the updated multicast trees. The SD-controller is managing the transmission time allocation for each SD-ONU and also manage the wavelength of SD-OLT. Further, the SD-OLT and SD-ONU mechanisms are given in Subsections C and D. Finally, the end-user device, it contains the media player and VR head mounted display (HMD) and the packet receiver. The Packet Receiver processes the packets received from optical access networks and reassembles them into a single video stream so that the media player and HMD can decode and playback the video stream.

C. SOFTWARE-DEFINED OPTICAL LINE TERMINAL (SD-OLT)
The SD-OLT allows centralized PON control and connects the PON system to the multi-access edge computing server on content delivery networks and illustrated in Figure 2. This system was designed to support the NG-EPON standard protocol and to enhance the SDN OpenFlow elements and protocols in order to provide multi point media access control (MAC) and SDN functionality to the EPON system. This combination provides the management infrastructure needed for the NG-EPON system's dynamic network resource provisioning. The SD-OLT has two L-OLT connected to the star coupler (SC) and SD-ONUs. Each L-OLT is configured with a fixed wavelength transceiver that operates on a distinct wavelength pair and is connected to its own operation, administration, and management (OAM) and MAC control clients. Moreover, this OAM manages the ALR functionalities and MAC maintains the standard MPCP mechanism, which is enhanced by OpenFlow signaling to augment the PON system with additional functionality that is not provided by the MPCP. The SDN controller orchestrates the L-OLTs, OAMs, and MAC control clients to distribute wavelength and link rate tuning signaling to SD-ONUs. Figure 4 shows the SD-OLT with OpenFlow switch architecture. SD-OLT has five major submodules as 1) OpenFlow Switch (OF); 2) OLT buffer; 3) ISP information interface; 4) Multicast Management; and 5) DWBA, all of which are described as follows. The key function of these submodules is to mark and multicast the most popular channels or videos, reducing server resource consumption.

1) OPENFLOW SWITCHE (OF-SWITCH)
An OF switch is responsible for data forwarding according to the flow table entries. Through the southbound interfaces, the matching rules and actions/instructions are provided by the SD-controller. The collection of matching rules is designed to recognize and extract the TI-P2P (360°) live streaming packet from the other packets. These process outputs include some essential information such as the channel ID, the channel title, the source/destination address, the user datagram protocol (TCP/UDP) source(src) and the destination(dst) port. Once the incoming packet is analyzed, the SD-OLT (OLT-table) verifies the packet information to determine if the livestreaming packet belongs to the most popular channel. If the packet is the most popular video channel chunk, the stream is placed in the GLID-SCB queue. When a packet (i.e. the most popular video channel chunk) is sent from the GLID-SCB queue, each SD-ONU can receive and store the stream in its buffer. Further, the OF switch will also be used to reroute localized packet mechanism peer-stream requests within the same ISP, thereby helping to reduce inter domain traffic costs. Finally, SRAM is utilized to store data-plane functions.

2) SD-OLT BUFFER
This buffer maintains the video chunk of the frequently play most popular channel. In live streaming, OLT need not maintain all the videos in the buffer, unlike Video on demand (VoD). Caching the complete live channel is impractical, as the video chunk is inaccessible until the content provider generates it. The cached video channel chunk is the portion that peers still have to play. Therefore, the amount of buffer to preserve or keep the most popular video channel chunks is not substantial compared to VoD. Hence, in this paper, we introduced the media and cache server separately, because the buffer size is low. The cache server maintains the most popular video and contents temporarily for the other applications. It also reduces the inter ISP traffic.

3) ISP INTERFACE
This interface maintains the important information, including storing vital information, such as routing table, topologies, traffic patterns, roundtrip time (RTT) among peers, and available bandwidth. This information assists network operators to manage the OpenFlow switch via the SD-controller to modify the matching rules as needed.

4) MULTICAST MANAGEMENT
This functionality is maintain the multicast information manager information such as user information (UI), multicast group (MG), Multicast Traffic sharing rate (MTSR), Traffic Sharing Density (MTSD) and Multiact local peer and client (MLPC). The UI function is maintaining the user database like user Id, SD-ONU-GLID and user login information based on the OpenFlow flow table data entries. This MG group is managed the SD-ONU GLID, multicast membership information, session information (JOIN or LEAVE the session) and identifier of the MG group of peers which receiving streaming data. (Note that the MG group is configured the SD-OLT domain). Therefore, the multicast group of the video contents and communicates to the multicast peer using a UDP protocol in the application layer. Accordingly, multicast packets must be handled efficiently in SD-OLT and SD-ONU. To transfer packets through a multicast session, the SD-OLT assigns al group link identifier (GLID) or multicast LLID to the MG in the downstream transmission. Next the MTSR represents the number of SD-ONUs or GLID (Group of SD-ONUs based on IEEE 802.3ca policy) Join the multicast session. Moreover, the MTSR is generally used in the measurement of multicast efficiency (effective bandwidth allocation and provide effective throughput). For example, the number of watchers, who are watching the same television channel or videos Live IPTV, is the MTSR of that channel (multicast session). The MTSD functionality represents the normalized multicast service efficiency that can be provided by session   [33]. Finally, MLPC is based on the IP address of the multicast local peer that is providing streaming data to the peer clients. The MG module of the multicast information manager (MIM) monitors all incoming requests for live video and regulates the start and termination of the user multicast session. Simultaneously, it manages all active users' information in multiple tables, such as MulticastSessions Channelnfo, ViewerInfo and, UserInfo, as shown in Figure  5. The MG group table contains the service group id (S_G_ID), master ONU index, subscriber, multicast id and the channel ONU (C_ONU) i.e it maintains specific viewers for particular C-ONU information. The MulticastSessions table stores information about currently running multicast sessions, such as the Group Service Id (which is dynamically generated by Media Server), Channel Info (C info), multicast address Id, and the Channel _ONU (C_ONU) associated with the service. The channelInfo table stores the necessary information about all active channels or videos being viewed within the service region surrounded by the core network, such as the channel id (C_id), the total number of viewers, the total number of cells (T_Cell), and the status of streaming type status, such as unicast or multicast. The ViewersInfo on SD-ONU table contains information about viewers who have viewed a specific video within a specific C_ONU coverage area, as defined by the channel. The UsersInfo table contains the user_id (ULID-User Linke Identifier), the user's mobile number, the user's C_ONU, and the ChanelInfo (C_Info) being watched.

5) DWBA
This module performs the dynamically assign wavelength and bandwidth allocation. SDN controller dynamically manages the wavelength assignment. Moreover, the P2P-TI DBA inter and intra traffic scheduling techniques are used. The details of the proposed DWBA scheme can be found in section III.
Moreover, the SD-OLT architecture shows the incoming packet of the pipeline processing to the internet. When user use P2P live-streaming software, they become a streaming node and join the network. The first step is to download the peer-list root server's channel lists. The user then registers with the peer-list root server and requests an initial list of peers who are also watching the channel. The peer can then communicate with other peers on the list to obtain more lists to add to the current one. In this way, the root server's signaling overhead is reduced. When an incoming packet enters the OF switch, it is processed against the entries in Flow Table1 (see Figure.4). As soon as a flow entry is discovered, the packet is directed to Flow Table2 by the instructions set. These steps are repeated until the final flow table. In our design, Flow Table1 distinguishes and marks P2P-TI live-streaming packets based on their source, destination, and TCP/UDP src/dst ports. The matched P2P live-streaming packet is then forwarded to Flow Table2 for further processing. The flow entry in Flow Table2 extracts packet metadata such as channel name, channel ID, bitrate, etc. Flow Table3 modifies the packet and updates the metadata after receiving it from Flow Table2. If the packet has successfully passed through the entire pipeline process, it is forwarded to the SCB_GLID port and stored in the SD-OLT buffer; if not, it is queued in the Tx queue. For instance, if the peer-to-peer live-streaming protocol evolves, new matching rules can easily be updated via a controller.

D. SOFTWARE-DEFINED-OPTICAL NETWORK UNIT (SD-ONU)
The SD-ONUs link subscribers to the network through its user network interface (UNI) and every L-ONU is outfitted with a tunable transceiver that can be tuned to any of the twowavelength pairs that make up the PON system. The MAC control client is responsible for the administration of the PON system on a basic level, such as registering the SD-ONU with the SD-OLT, receiving GATEs, and issuing REPORTs in accordance with the MPCP protocol standard.
As is shown in Figure 6, the SD-ONU logical architecture in this paper consists of three submodules: OpenFlow switch (OpenFlow Agent and Flow Table), Tunable transceiver, Intrauplink scheduler and ONU buffer. These modules are used to classified and process the live P2P-TI packets on ONU based on the network traffic flows i.e., Src/Dest TCP port that the peers are using, protocol type and the IP address. The main contribution of this submodules:  1. These submodules help to identify P2P-TI packets and locate the most-requested P2P live streaming channel. 2. It seeks to reduce the P2P-TI packet delay time by serving the request locally, which reduces the overall delay time.

1) OPENFLOW SWITCH (OF-SWITCH)
The OF-Switch is consist of two main parts: 1) OpenFLow Agent and 2) Flow

3) ONU-BUFFER
It is possible to keep only the most popular videos or large segment of videos in a ONU buffer map, which is located at each ONU. When it comes to our architecture, the buffer map performs a similar purpose to the caching scheme. It will supply consumers with the segments that are required for streaming a video on the internet. Moreover, ONU can also cache up to a few minutes of video chunks for popular live streaming. These segments may have just been played or will be in a few minutes.

4) INTRAUPLINK-PON SCHEDULER
It signifies that the requested segments are in another buffer map or that the other users are in a different subnet, but they are still in the same PON if the requested segments are tagged as intra-PON traffic. The four different queues must be scheduled for the ONU according to our proposed scheduling approach. P2P-TI live-streaming traffic is given a higher priority in this article than AF traffic, on the basis of our assumptions. Contrary to popular belief (one buffer for all traffic types), But the proposed intrauplink scheduler provides a buffer for each traffic type [15]. The scheduler attempts to serve first prioritizes in EF traffic, followed by P2P-TI live streaming, AF and BE traffic.

VI. PROPOSED SD-TI-DWBA
The QoS aware resource managements in EPON are widely investigated in the literature [34]. In addition, the SDN wavelength, link rate and overall orchestration mechanism have been investigated the literature [29]. SDN based tactile internet resource provision in smart hospitals using NG-PON2 are studied in Ref [35]. The SD-TI-DBWA mechanism is used in S-OLT in our proposed architecture to dynamically allocate bandwidth across two pairs of wavelengths, regardless of the time windows/slots for upstream data transmission at each SD-ONU. Because the proposed technique employs an offline DBWA approach, the SD-OLT will begin allocating wavelengths and bandwidth after receiving REPORT messages from all SD-ONUs. In other words, each SD-ONU may transmit only at the wavelength(s) assigned to it during the allotted time window to avoid optical distribution network collisions (ODN). After collecting and extracting all REPORT messages, the S-OLT will acquire information on each SD-queues, ONU's bandwidth requirements, and so on, via the SDN controller.
In addition, according to [36], the DWBA can be divided into three categories:1) single scheduling (SSD), 2) multi scheduling (MSD) and 3) wavelength-agile (WA) schemes. The DWBA is responsible for allocating different multiple available channels and timeslots to various SD-ONUs. The WA domain was chosen for this proposed mechanism because studies show that user traffic behavior in access networks varies according to busy hours and high or lowpriority traffic [37]. WA-PON enables complete wavelength assignment and sharing flexibility, as any ONU can be assigned an unlimited number of wavelengths at any given time, and many ONUs can transmit concurrently as long as their allotted wavelengths do not overlap. For example, during the day, some SD-ONU(s) providing tactile services, i.e., SD-ONU(s) serving office buildings, may have higher bandwidth than other SD-ONU(s) serving residential customers. Additionally, by adopting SDN in SD-OLT, network operators can alter the wavelength and bandwidth allotment to each SD-ONU adaptively in accordance with the service level agreement (SLA).
The SD controlled dynamic resource allocation scheme to manage the wavelength, link rate, and time of an NG-EPON in order to minimize the low latency of the system. Moreover, the aim of this proposed SD-TI-DWBA scheme which supports all types of haptic feedback and AR/VR HD,4K video streaming and aims to achieve low latency guarantees for tactile and non-tactile services. The proposed SD-TI-DWBA consists of three modules: 1). SD-DWBA operations, PON activation/deactivation, wavelength, link-rate management and QoS management in the network. 2). SD-time management, DWBA scheduling mechanism, it manages the Inter/Intra SD-ONU traffic scheduling and transmission time allocation, grant sizing for each CoS of each SD-ONU and is responsible for grant scheduling the upstream channel. 3). The wavelength-agile (WA) techniques, WA-PON enables complete wavelength assignment and sharing flexibility, whereby any number of wavelengths could be allocated to any ONU at a certain time, and the number of ONUs can transmit concurrently as long as their assigned wavelengths do not overlap, as illustrated in Figure 7. Notice in this figure that ONU6 has much more data to transmit than ONU1 and ONU8, so we assign ONU6 two wavelengths and ONU1 and ONU8 each one. In this case, ONU6, ONU1 and ONU8 can transmit simultaneously, at the same time ONU6 and ONU7 can share the same wavelengths [36,38]. Table 2 summarizes parameter specifications for the SD-TI-DWBA mechanism.

A. SD-TI-DWBA
1. The SDN manages the wavelength, transmission linkrate, and time of an NG-EPON in order to reduce the latency and energy of the system. 2. The SDN process the PON system activation and monitoring, which is initiating the registration procedure by activating the L-OLT with the available link-rate in the SD-OLT. 3. Initiates the OAM process detection in L-OLT and registers discovered ONUs with the SDN controller. Further, the SDN controller registers the newly added SD-ONU in its database. 4. Finally, the SD-OLT begins the DBA process based on the ONU REPORT and OLT GATE message. Finally, the SDN controller monitors the overall PON system traffic and manages the active L-OLT network load and adjusts the link rate. As mention above, to manage the wavelength and transmission link-rate (TLR), the SDN controllers active L-OLT traffic load and SD-ONU buffer status periodically collect and store the statistical information in their database. This statistical data is utilized to determine the optimal wavelength and link rate configuration. However, traffic circumstances must be dynamically modified by system provisioning. The SDN controller must thus determine the required number of active transceivers based on traffic load.
where ONU load, N, j, and Wthroughput denote the required ONU data, the number of registered SD-ONUs, the number of recognized traffic types by the SD-ONU, and the maximum available throughput with the maximum link rate, respectively. After the active transceiver is configured, the SDN controller must be set the active L-OLT transmission line rates ( ). We use the link rate (LR) as minimum 1 Gbps and a maximum of 25 Gbps.
where , 1G, 25G, denoted the bank of transceiver transmission link-rate, and represent the threshold of the requested downstream bandwidth of all ONUs on that particular wavelength, respectively. And also, the SDN controller maintained the transmission time allocation on each SD-ONU from the SD-OLT using the DBA agent. Figure 8 illustrates the transceivers (L-OLT) and link-rate configuration process.

B. SD-TIME MANAGEMENT WITH DWBA SCHEDULING MECHANISM
In the proposed SD-DWBA scheme we assume that TI traffic is wavelength 1 and NTI traffic wavelength 2 . However, both 1 and 2 , are eventually shared for all traffic as needed. Here, we use the wavelength agile approach to reduce the latency. Thus, each ONU contains the aggregate buffers' dwelling units for TI (∑ , =1 ) and NTI (∑ =1 ) traffic in its REPORT message. Consequently, the SD-ONU allocated the intertraffic and intra traffic scheduling to meet the QoS services on each CoS class.
The SDN controller manages the transmission times allocation for each SD-ONU via the DWBA agent in the SD-  OLT in the proposed multicast architecture. Further, DWBA dynamically assigns bandwidth to every SD-ONU in accordance with its requests. In the proposed SD-DWBA, after receiving all REPORT messages from each ONU, the SD-OLT computes bandwidth allocation. Then, the SD-OLT implements proposed SD-TI-DWBA, which is calculated the total available bandwidth ( ) for each SD-ONU and maximum transmission window ( ).
where represents the total available bandwidth of _ in bytes, represents the transmission transceiver link-rate (speed) of the NG-EPON, is the default polling cycle time , is the guard time, and the control message length is 512 bits (64 bytes) in the NG-EPON system. The DBA assign the maximum transmission window limit.

= .
(4) varies depending on the traffic type (EF, AF, TI-P2P and BE,), with higher priority traffic taking precedence over lower priority traffic.
The proposed SD-TI-DWBA mechanism as follows: The SD-TI-DWBA traffic priority and transmission time slot management operations are illustrated in Figure 9. Here there are two active wavelengths channel#1 and channel#2) and two active SD-ONUs (SD-ONU#1 and SD-ONU#2). SD-ONU#1 is connected channel#1 and SD-ONU#2 connected channel#2 in one cycle. An all SD-ONU must send a REPORT message to the OLT in order to be assigned an upstream transmission timeslot. When OLT receives REPORT messages from all SD-ONUs, it obtains buffer occupancy information for each SD-ONU. This information is used by OLT to calculate and allocate bandwidth for each traffic type (i.e., EF, AF, TI-P2P, and BE). This polling process is used to determine the state of each SD-ONUs buffer. Following receipt of all REPORT messages, the DBA may compute the upstream transmission timeslot to be assigned to each SD-ONU. As a result, once the calculation is complete, the OLT sends GATE messages to all SD-ONUs via the corresponding channel. The GATE messages include upstream transmission timeslots that the SD-ONUs can use to send their buffered packets from subscribers. The SD-TI-DWBA mechanism has the advantage of reducing idle time, delay, and OLT overhead, thereby improving the SD-NG-EPON system's performance.
The SD-OLT DBA scheme is illustrated in Figure 10. At first, the OLT collects each ONU's TI request bandwidth and then determines the grant bandwidth , for TI traffic for cycle + 1, , +1 , depending on the SLA for each SD-ONU sum of TI and NTI traffic bandwidth request, ∑ , =1 and ∑ =1 , Then, we first compute the can be calculated as ∑ =1 , is bigger than . The formula is illustrated as follows: Otherwise, to avoid delays in high-priority traffic packets, the SD-OLT will satisfy each ONU's TI bandwidth request and then calculate the remaining available bandwidth , . The can be expressed in the following manner: Therefore, adjusts the size of to find the real amount of bandwidth allotted for tactile traffic. The OLT allocation the remaining bandwidth for non-TI queues according to the SLA. ≤ (7)

C. WAVELENGTH AGILE-MECHANISM
WA-PON is a hybrid EPON comprised of MSD and SSD PONs. In WA-EPON, the OLT can allocate a single wavelength or multiple wavelengths to different ONUs concurrently based on their demands. If an ONU wants to broadcast at multiple wavelengths, tunable lasers should be used. In WA-EPON, the OLT grants access to a multiwavelength ONU for the purpose of transmitting its queued data. To schedule transmission effectively, an arbitrator is required to avoid collisions in the upstream direction. Further, the WA-PON architecture and operation is shown in Figure  7. Figure 11 shows the NG-EPON wavelength agile interchannel or adaptive multichannel wavelength sharing mechanism. The SDN controller monitors the overall traffic and activates and deactivate the L-OLT based on the traffic conditions. The SDN is more intelligently help to the wavelength agile management based on this timeslot allocation. Figure 11(a) when = , TI and NTI queues are transmitted on his dedicated wavelength. Figure  11(b) when > , the WA-agile is performed in this stage, the TI queues is transmitted on both wavelengths and NTI queues are transmitted one wavelength. The same approach for the Figure 11(c), when > , transmitting over the tactile wavelength from the non-tactile queue up to the predefined threshold value. Therefore, the WA-agile mechanism enhances the overall system QoS performance.

V. PERFORMANCE EVALUATION
In this section, we analyze the proposed mechanism in terms of the mean packet delay, jitter, queue length, throughput, packet loss ratio. The proposed SD-TI-DWBA was compared with the system that uses Non-TI-DWBA. The proposed mechanism was modeled using an OPNET simulator with 64 SD-ONUs and an SD-OLT with 2 L-OLTs. The efficient downstream/upstream channel rate between SD-OLT and SD-ONU is dynamically assigned 1 or 25 Gbps. The distances from the SD-ONUs to the SD-OLT are assumed to be between 10 to 20 km, and each ONU has a 10 MB buffer. The maximum transmission cycles are 1.0 ms. The network traffic model for AF, TI and BE is based on self-similarity and longrange dependence. The recent studies in [39] have assumed that tactile internet traffic is Pareto distributed. This model will originate with high-burst AF, TI and BE traffic with a Hurst parameter of 0.7 and a packet size uniform (AF and BE), Pareto (TI-P2P-live streaming traffic) distributed between 64   and 1,518ytes. The high-priority traffic (EF) model uses the Poisson distribution with a fixed packet size (70 bytes). Moreover, the video-transmission rate of TI-P2P live streaming is set to 500 kbps and 1 Mbps from scenarios S1-S6. The simulation parameters are summarized in Table 3. Four scenarios shown in Table 4 are designed and analyzed with various EF, AF, TI, and BE service proportions to show the effectiveness of the high-priority traffic management.

A. MEAN PACKET DELAY
When packets arrive at the ONU at random times, the mean packet delay occurs. Each packet must wait for time periods in the upstream direction before transmitting. The packet delay is consists of the polling delay, the granting delay, and the queuing delay [40]. However, for TI-P2P live-streaming, the mean packet delay includes the foregoing variables plus application-specific processing delay. This processing time varies with hardware and flow-table design. Our simulation used a 0.01 ms processing delay. Figure 12 shows the mean packet delays for EF, AF, TI-P2P live streaming different scenarios versus the offered load with 64 ONUs, respectively. The simulation results show that with intra traffic, the P2P-live streaming delay with a 1 Mbps video transmission rate improves dramatically. It can be seen that by utilizing the OF switch to regulate P2P live-streaming traffic, the mean packet delay for all packet classes can be decreased in all scenarios. By deploying the SD-NG-EPON architecture and buffer (cache) at the ONU, network operators can reduce costly inter-ISP traffic while simultaneously improving streaming quality, as streaming arrives locally. The proposed architecture with the proposed SD-TI-DWBA reduced the mean packet delay for TI-P2P live-streaming by 20.62% when the offered load was 100% in the S3 scenario. Additionally, despite the high P2P live-streaming traffic rate (S3), the proposed architecture with a mean packet delay of less than 3.5ms for P2P livestreaming maintains a video transmission rate of 1 Mbps while meeting the EF and AF traffic requirements.

B. TI-P2P JITTER
The TI-P2P delay variance, or TI-P2P jitter, can be calculated as 2     number of received TI-P2P packets. Figure 13 shows the TI-P2P jitter versus the offered load performance for SD-TI-DWBA under the different scenarios with 1.0 cycle times. The TI-P2P jitter has a trend similar to that of the TI-P2P delay. As the offered load increases, the variance of the TI-P2P mean packet delay increases, resulting in an increase in TI-P2P jitter to approximately 0.11 to 0.13 for all scenarios. Hence, due to OF switch traffic localization methods, the proposed SD-TI-DWBA architecture offers better jitter performance when traffic load is less than 100%. As a result, it can be inferred that traffic OF switch localization has a considerable effect on the performance of TI-P2P jitter.

C. SYSTEM THROUGHPUT
The system throughput is the sum of the data rates given to all network terminals. The system throughput in this proposed architecture is the sum of transmission between the ONU and OLT communication and local traffic throughput. The local traffic refers to traffic that is handled by ONUs without reference to feeder fiber or OLT. The OLT transmits TI-P2P channel data one time, and the ONU is liable for multicasting the data to users in the proposed design. Figure 14 show the system throughput SD-TI-DWBA and NT-DWBA in different scenarios versus the offered load with 500kbps and 1 Mbps video-transmission rate, respectively with 64 ONUs. The system throughput is computed by multiplying the PON line rate (i.e., 2×25 Gb∕s) TImanagement and encapsulation overheads. When the share of local traffic increases, the system's throughput increases proportionately. The system throughput of all proposed scenarios at high loads is more than 1 Gb∕s. This is because the ONUs handles a portion of the traffic locally via OF Switch and buffer cache, and they also execute multicasting. As a result, increasing the video transmission rate results in an increase in system throughput, as certain peers exchange streams locally without transmitting them to the OLT or the Internet. As a result, the suggested architecture is scalable, as the video transmission rate grows, the system throughput increases as well.

D. PACKET LOSS RATIO
The ONU buffer size constraint causes packet loss. Placing packets in the queue causes the queue to fill, forcing packet dropping. Figure 15 compare the average BE packet loss probability in the different scenarios versus the offered load with 64 ONUs. The packet loss probability for high priority traffic such as EF, AF, and TI-P2P live streaming is zero in all scenarios. The packet losses of NT-DWBA for the offered load below 80% in all scenarios are zero. On the other hand, when the offered load exceeds 90%, SD-TI-DWBA scenarios S1, S2, S3, S4, S5, and S6 begin to drop packets, and the packet drop ratio reduces when compared to NT-DWBA and SD-TI-DWBA, because of proposed mechanism manages a portion of the request locally. Moreover, the BE packet losses of the SD-TI-DWBA and NT-DWBA in the case of high traffic loads were different in all scenarios, which indicates that the proposed architecture will not affect the packet loss.

VI. CONCLUSION
In this paper, we have presented the 5G enabled tactile internet-based 360° VR video streaming architecture. we have proposed a new multicasting mechanism to deliver 360° TI-P2P live streaming traffic over a MEC-enabled software-defined-NG-EPON architecture to enhance the overall QoSs. IEEE 802.3ca based, a unique GLID is assigned to a group of intra TI-P2P channels for building a multicast mechanism and handling user TI-P2P requests with streaming protocols. The proposed architecture SDN multicasting controller and the OF switch-based redesigned at the OLT and ONU. The redesigned OF switch-based ONU is responsible for multicasting the received frames from the OLT to the end-user. It can provide the ISP a better softwarebased traffic analysis and centralized control that can improve the overall QoS and bandwidth utilization. The proposed multicasting SDN controller module at the OLT, which aims to localize the overlay P2P live-streaming network by saving the most viewed video channel chunk has the potential to optimize intra-ISP traffic, and hence the expensive inter-ISP traffic. Moreover, the SDN controller orchestrates dynamic resource allocation to resolve conflicts between P2P live streaming, active wavelength management, connection rates, timeslot allocations, and QoS performance. Finally, the SD-TI-DWBA assigns appropriate timeslots and wavelength agile to packets based on their priority to ensure that QoS requirements are met. Simulation