Journal of Electronic Science and Technology  2019, Vol.17 Issue (3): 242-251   DOI: 10.11989/JEST.1674-862X.71018158 PDF
http://dx.doi.org/10.11989/JEST.1674-862X.71018158
51K

Article

Hui-Kai Su, Jian-Ting Pan, Kim-Joan Chen, Marek R. Ogiela, Hsing-Chung Chen
Real-Time Streaming Relay Mechanism for P2P Conferences on Hierarchical Overlay Networks
Journal of Electronic Science and Technology, 2019, 17(3): 242-251
http://dx.doi.org/10.11989/JEST.1674-862X.71018158

Article History

revised April. 18, 2017
Real-Time Streaming Relay Mechanism for P2P Conferences on Hierarchical Overlay Networks
Hui-Kai Su, Jian-Ting Pan, Kim-Joan Chen, Marek R. Ogiela, Hsing-Chung Chen
H.-K. Su is with the Department of Electrical Engineering, National Formosa University, Yunlin 632 (e-mail: hksu@nfu.edu.tw);
J.-T. Pan is with DrayTec Corporation, Hsinchu 30041;
K.-J. Chen is with the Department of Electrical Engineering, National Chung Cheng University, Chiayi 621;
M. R. Ogiela is with Akademia Górniczo-Hutnicza University of Science and Technology, Krakow 30-059 (e-mail: mogiela@agh.edu.pl);
H.-C. Chen is with the Department of Computer Science and Information Engineering and the Department of Bioinformatics and Medical Engineering, Asia University, Taichung 41354, and also with the Department of Medical Research, China Medical University, Taichung 40402 (e-mail: cdma2000@asia.edu.tw)
This work was partly supported by MOST under Grants No. 103-2221-E-150-020 and No. 104-2221-E-468-002 and Asia University and China Medical University Hospital, China Medical University under Grant No. ASIA-105-CMUH-04
Abstract: A peer-to-peer (P2P) multimedia conferencing service is operating that users share their resources to each other on the Internet. It can solve the problem in the centralized conferencing architecture, such as the centralized loading, single point error, and expensive infrastructure. However, P2P networks have the problem that a peer has a difference between the physical location and logical location in the overlay network. In the viewpoint of P2P networks, the nearest conference resource may be far away geographically. The P2P-session initiation protocol (P2P-SIP) multimedia conference is to construct an application-based logical multicast network efficiently according to physical network information. Thus, this paper proposes a real-time streaming relay mechanism for P2P conferences on hierarchical overlay networks. The real-time streaming relay mechanism can improve the transportation efficiency of conferencing stream exchange well based on the application-layer multicast (ALM) structure and the hierarchical overlay networks.
Key words: Application-layer conferences    hierarchical networks    multimedia conferences    path recovery    peer-to-peer (P2P)    session initiation protocol (SIP)
1. Introduction

In a traditional centralized multimedia conferencing structure, there are many issues, such as expensive infrastructure, weak scalability, and single point error. The above-mentioned issues will make the service provider spend a lot of cost for system establishment and maintenance when the user’s demand becomes more and more.

In the recent years, there are many researches focusing on the peer-to-peer (P2P) system framework. The user sharing the resources of conferences with each other in the distributed architecture can reduce the cost of infrastructure. But a user’s virtual location in the P2P overlay network differs from the user’s physical location, which makes the network produce a redundant route when the conference is working. The geographical location of peers cannot be recognized by the traditional P2P system well. In the viewpoint of the P2P network, the nearest conference resource may be far away geographically. Therefore, the P2P conference system can not build a fit multimedia application-layer multicast (ALM) tree when the system is shortage of physical network information in a distributed environment. For example, when both peers in the same local area network communicate with each other, the conferencing streams may be forwarded to the peer that is sharing the conference resource in another country or continent, and then redirected back. In the situation, if the conference participant can discovery the nearest peer sharing conference resource, the system performance and communications quality would be improved.

2. Related Works

The Internet Engineering Task Force (IETF) established many working groups to research and present the standards about multimedia conferences. It also formulates many Request For Comments (RFC) documents, e.g. RFC 4353[1], RFC 4579[2], and RFC 5850[3].

The P2P-session initiation protocal (P2P-SIP) is one of the important references of P2P real-time applications. It combines with the P2P distributed structure and the traditional centralized SIP client-server structure. It uses an overlay network to provide the service by which the full-distributed resource register and message transmit with SIP. It has some advantages, i.e. decentralization, low cost, high toleration, and high scalability, to improve the SIP system’s problems. P2P-SIP uses the SIP-over-P2P structure. It uses the P2P protocol to provide register and location to the SIP protocol. The REsource LOcation And Discovery (RELOAD) Base Protocol[4] was proposed in 2014. Moreover, some studies deal with the secured digital content and multimedia quality of service (QoS)[5]-[9].

Because the current network environment does not support the Internet Protocol (IP) layer multicast function universally, ALM would be the main way that supports conferences to relay the streaming. The current studies about building the fit ALM tree for the P2P architecture are less.

3. P2P-SIP Conference System 3.1 Hierarchical Overlay Network

The hierarchical overlay network is constructed by peers. After authorizing, peers can join the overlay network. The P2P-SIP conference system is based on the overlay network. Each peer will create the neighbor information, and manage P2P routing information by an overlay network routing algorithm. Each peer provides computing and storage resources to reserve data, which should be kept online. The previous study[10] brings up the idea that IP prefix is used to divide a P2P network into several local overlay networks. This idea can make peers to construct local overlay networks according to their physical networks geographically. The mechanism can reduce the redundant route. The previous study[11] drafts the concept. The local overlay network transmits messages to other local overlays via the common overlay, which is shown in Fig. 1.

 Fig. 1 P2P-SIP conference system with the hierarchical overlay network.

The common overlay is constructed by several super nodes. In our system, the super node is defined that a peer can provide a large, high reliable, and high available conference resource. The super node joins the local overlay belonging to him and joins the common overlay simultaneously. The super node is one of the local overlay gateways. It handles the communications between the local overlay and the common overlay in the conference control plane.

A peer can share resources according to the peer’s capacity to another peer when it becomes a P2P focus or a P2P mixer. The P2P focus provides the functionalities of the conference control plane that manages and handles the P2P conference operation. It makes session initiation and manages the resources of conferences. The P2P mixer provides the functionalities of the conference data plane. It is used to mix voice data and video data, and deliver the mixed streams to each participant in a conference.

In order to provide reliable and high quality P2P communications, how to select a suitable peer to provide P2P resources is very significant. In a general environment, peers with high bandwidth capacity and low transmission latency may provide a high quality and high reliability transportation service[12],[13]. However, the heavy loading of the P2P network is incurred by measuring bandwidth frequently, and the conference set-up time will be increased. Thus, the predict delay in this paper is based on the landmark predictive mechanism[14]. In the same local overlay network, the peers are located closely, and the delay is not considered by peers. The super nodes on each local overlay networks will measure the round trip time (RTT) to all landmarks periodically and get landmark vectors, which is shown in Fig. 2.

 Fig. 2 Super nodes measure RTT by landmarks.
3.2 Priority Weight

In order to provide a reliable and high-quality P2P conference service, how to identify the level of peers is very important. In our system, the priority weight metric is defined according to the delay, connectivity, or stability.

3.2.1 Peer Delay

Based on landmarks and the churn-resilient protocol (CRP)[15], the formula which can predict the delay between two local overlay networks is proposed. Two different regions overlay networks are signed by x and y, $d_i^{\,x}$ is the delay between the local overlay network x and landmark i, and $d_i^{\,y}$ is the delay between the local overlay network y and landmark i. The delay between the local overlay network x and the local overlay network y can be expressed as

 $d(x,y)\; = \;\sqrt {\small\sum\limits_{i = 1}^n {{{(d_i^{\,x} - d_i^{\,y})}^2}} }$

where n is the number of landmarks.

And $d(x,y)$ is defined and normalized to calculate the priority weight. The standardized formula is as follows:

 ${d_n}(x,y)\; = \;\frac{{300\;{\rm{ ms}}}}{{d(x,y)}}.$

When ${d_n}(x,y)\!>\!1$ indicates that the greater value expresses the lower delay, while this way can not be used if ${d_n}(x,y)\!\!\leq \!\!1$ .

The one-way delay for audio communications should not exceed 150 ms which was defined by the International Telecommunications Union (ITU). The delay that we measured is RTT, so the double one-way delay is to normalize our value.

3.2.2 Peer Connectivity

In our system, all peers will measure and predict their available bandwidth periodically. How to predict the available bandwidth is not discussed in this paper. The connectivity metric of peer i is defined as below.

 ${c_i} = \;\frac{{{\rm{A\;\!\!va\;\!\,\!ila\;\!\,\!ble\;\;ba\;\!\,\!ndwidth}}}}{{{\rm{T\!he\;\;ba\;\!\,\!ndwidt\;\!\!h\;\;whic\;\!\,\!h\;\;o\;\!\,\!ne\;\;s\;\,\!\!t\;\!\!r\;\!\!e\;\,\!\!a\;\!\,\!ming\;ne\;\,\!\!e\;\,\!\!ds}}}}$

where ${c_i}$ is the number of connections that the peer can support. Its larger value expresses more corrections once ${c_i} \!\!>\! \!1\;$ and it is unable to establish a connection for ${c_i}\!\! \leq \!\!1$ . In this paper, we evaluate the peer connectivity by that how many streams can be handled by a peer. The bandwidth of one conference streaming would be calculated.

Moreover, the capacity of the local overlay network can also be evaluated by

 ${C_j} = \small\sum\limits_{i = 1}^n {{c_i}}$

where j represents the jth local overlay network, and n is the number of peers in the local overlay network.

3.2.3 Peer Stability

Peer stability is an important issue in a P2P network, because the churn will occur when the peer stability is too low. The peer may join or leave the P2P network anytime and repeatedly. The system will lead to be crashed or unstable. The phenomenon is called “churn”. In general cases, the stable peers are disposed on the top layer of the streaming multicast tree to reduce the effect of churn. The child peers will lose the parent’s downstream data when the parent’s peer leaves.

It is a very important issue to design the real-time relay mechanism with the balance between the delay and bandwidth when building the ALM tree. In this paper, we use the priority weight of peer stability to solve this problem. The formula about the priority weight of connections is shown as

 ${\rm{PW}} =\text{α} d(x,y) + (1 -\text{α}){C_y},\;0 \!\!\leq \!\!\text{α} \!\!\leq \!\!1$

where $d(x,y)$ is the P2P delay between two peers, ${C_y}$ is the peer connectivity in the ALM tree, $\text{α}$ is the weighting parameter of the system, and it can keep the balance between the delay and connectivity. The value of $\text{α}$ is between 0 and 1.

3.3 ALM Tree Policy

Our environment is a hierarchical overlay network, so we design the peer selection mechanism according to the characteristics of the environment.

The figure of multicast tree node selection strategies is shown as Fig. 3. The mixer is the root of the multicast tree for delivering the conference streams to each participant. In order to limit the multicast end-to-end delay, the depth of the ALM tree would be controlled.

 Fig. 3 Multicast tree node selection strategies.

In Fig. 3, each peer is colored and identified by different local overlays. For example, five peers colored by blue belong to the same local overlay named Local Overlay 1. There are three peers in Local Overlay 2. Other peers are identified in the same way.

All peers in the ALM tree are the participants in the same conference. One of the powerful peers would be selected to a P2P mixer. We observe the ALM tree by each peer. Besides receiving and playing the mixed audio and video streams, the parent peer would also relay the stream data to his child peers. Generally, the ALM tree may be built randomly. In Fig. 4 (a), the mixer delivers the mixed stream to other local overlays first. However, due to the connectivity capacity of the mixer, the mixer cannot deliver the mixed stream to all local overlays on the first level of the ALM tree. In the situation, many overlay networks may be cascaded into the ALM tree. Finally, the depth of the ALM tree is large, and the high end-to-end delay of multicast is incurred.

 Fig. 4 General strategy of multicast tree node selection: (a) detailed and (b) simplified diagrams.

In our ALM tree policy, we will select one mixer from the local overlay whose connectivity ${C_j}$ is larger than a threshold. In the first level of the ALM tree, the mixer would deliver the mixed stream to the peers in the same local overlay. In the scenario of Fig. 5 (a), each blue peer relays the mixed stream to other overlay networks. Finally, the depth of the ALM tree would be limited.

 Fig. 5 Proposed strategy of multicast tree node selection: (a) detailed and (b) simplified diagrams.

Moreover, the ALM tree can be simplified and represented as Fig. 4 (b) and Fig. 5 (b). According to the general strategy, we can easily observe that the depth of the ALM tree is ‘2’. Based on our ALM tree policy, the depth of the ALM tree is ‘1’ as shown in Fig. 5 (b). Therefore, our real-time streaming relay mechanism can achieve small end-to-end delay.

4. Design of Real-Time Streaming Relay Mechanism and Its Functionality Comparisons

The real-time streaming relay mechanism for P2P conferences can be achieved in a hierarchical overlay network. The creation processes of the P2P-SIP conferencing service we proposed are shown in Fig. 6. The details of the system operation and message sequence chart are shown and explained below.

 Fig. 6 P2P-SIP conferencing service creation process[11].

1) The first step is to set up the profile of the P2P conference service, the profile includes the number of members, conference member’s SIP-uniform resource identifier (SIP-URI), and some mechanisms.

2) The caller checks the location of all members by SIP-URI, and looks for the focus on the overlay network by using the P2P distributed hash table (DHT) algorithm, i.e. Chord algorithm in this paper.

3) After selecting the P2P focus, the conference initiator submits the SIP invite message to the focus to setup a conference. The invite message includes the INVITE-Contained List that contains the list of the conference members. The conference members are the participants that will be invited into this conference.

4) According to the INVITE-Contained List of SIP messages, the P2P focus searches the conference members in each local overlay. By using the enhanced P2P DHT algorithm, the IP address of conference participants would be looked up in the overlay networks. In the same time, the information of landmark vectors of each overlay network can be obtained by the focus. The details of landmark vectors are explained in subsection 3.2.1.

5) All information of the available bandwidth and stability is responded by the conference participants. The details are explained in subsection 3.2.3.

6) The focus searches the available mixer by the Chord algorithm in the overlay network according to our ALM tree policy. The feasible candidate of the P2P mixers will be selected.

7) The focus transmits the invite message of SIP to the mixer.

8) The mixer responds 200 OK to the focus when it agrees to serve the mixing task.

9) The focus transmits the invite message of SIP to other members who are presented in the list. The focus will give the information of the mixer to all members. In this step, the message includes the result that is according to subsection 3.2. Moreover, the mixer owns all participant information and the ALM tree would be constructed by the mixer.

10) After joining the conference, the conference members will transmit their audio and video streams to the mixer, and receive the mixed audio and video streams from the ALM tree. The conference is ongoing.

Finally, the functions comparisons among the other conference services are shown in Table 1. Our design for the real-time streaming relay mechanism and its functionality comparisons are better than the conference services in [1] and [10].

Table 1 Functions comparisons among other conference services
 Items Schemes SIP conference[1] Su et al.’s scheme[10] Our scheme Conference management architecture Centralized Distributed Distributed Conference mixer architecture Centralized Distributed Distributed Protocol SIP P2P-SIP P2P-SIP Conference streaming relay No No Yes Mixed streaming route Shortest path Shortest path Based on ALM Bandwidth efficiency Fair Fair Good
5. Conclusions

This paper proposed a real-time streaming relay mechanism for P2P conferencing services in a hierarchical overlay network. Based on the characteristics of the ALM structure and the hierarchical overlay network architecture, the priority weight was defined and the ALM tree policy was proposed. Our relay mechanism can improve the transportation efficiency of conferencing stream exchange, i.e. audio streams and video streams. On the first level of the ALM tree, the mixer delivers the mixed stream to the participants in the same local overlay, and then participants are responsible for relaying the mixed stream to the participants on the second level of the ALM tree and belong to different local overlays. On the remaining levels of the ALM tree, the stream would be relayed within the same local overlay. Thus, based on our real-time streaming relay mechanism, the end-to-end delay of P2P conferences would be bounded. Finally, the case study of the real-time streaming relay mechanism of P2P conferences was explained. In the future, the performance evaluation for huge peers will be simulated and analyzed.

References
 [1] J. Rosenberg. (February 2006). A framework for conferencing with the session initiation protocol (SIP). Network Working Group. [Online]. Available: https://tools.ietf.org/pdf/rfc4353.pdf [2] A. Johnston and O. Levin. (August 2006). Session initiation protocol (SIP) call control—conferencing for user agents. Network Working Group. [Online]. Available: http://buildbot.tools.ietf.org/pdf/rfc4579.pdf [3] R. Mahy, R. Sparks, J. Rosenberg, D. Petrie, and A. Johnston. (May 2010). A call control and multi-party usage framework for the session initiation protocol (SIP). Internet Engineering Task Force. [Online]. Available: http://www.hjp.at/doc/rfc/rfc5850.txt [4] C. Jennings, B. Lowekamp, E. Rescorla, S. Baset, and H. Schulzrinne. (January 2014). REsource LOcation And Discovery (RELOAD) Base Protocol. Internet Engineering Task Force. [Online]. Available: https://tools.ietf.org/rfc/rfc6940.txt [5] Y.-J. Lin, I.-C. Chang, H.-W. Yang, and C.-C. Yang, " Secure digital content distributing system on P2P networks,” Journal of Electronic Science and Technology, vol. 10, no. 3, pp. 263-268, 2012. [6] Z. Zhou, Y.-M. Mao, and Z.-Q. Jiang, " Calculation and analysis of destination buffer for multimedia service in mobile ad hoc network,” Journal of Electronic Science and Technology of China, vol. 3, no. 3, pp. 203-207, 2005. [7] H.-C. Chen, " A multi-issued tag key agreement with time constraint for homeland defense sub-department in NFC,” Journal of Network and Computer Applications, vol. 38, pp. 88-98, Feb. 2014. [8] H.-C. Chen, I. You, C.-E. Weng, C.-H. Cheng, and Y.-F. Huang, " A security gateway application for end-to-end M2M communications,” Computer Standards & Interfaces, vol. 44, pp. 85-93, Feb. 2016. [9] S.-S. Tseng, H.-C. Chen, L.-L. Hu, and Y.-T. Lin, " CBR-based negotiation RBAC model for enhancing ubiquitous resources management,” Intl. Journal of Information Management, vol. 37, no. 1, pp. 1539-1550, 2017. [10] H.-K. Su, C.-M. Wu, and W.-H. Yang, " Design of location-based hierarchical overlay network supporting P2PSIP conferencing service,” in Proc. of Intl. Conf. on Autonomic and Trusted Computing, Xi'an, 2010, pp. 26-29. [11] H.-K. Su, J.-T. Pan, C.-C. Lin, K.-J. Chen, and M.-T. Yang, " Design of a delay-bounded media relay mechanism for P2P conferencing services on hierarchical overlay networks,” in Proc. of 2013 Intl. Conf. on ICT Convergence, Jeju, 2013, pp. 228-231. [12] V. Sharma, R. Kumar, and N. Rathore, " Topological broadcasting using parameter sensitivity-based logical proximity graphs in coordinated ground-flying ad hoc networks,” Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications, vol. 6, no. 3, pp. 54-72, 2015. [13] G. Cicotti, L. Coppolino, S. D'Antonio, and L. Romano, " Runtime model checking for SLA compliance monitoring and QoS prediction,” Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications, vol. 6, no. 2, pp. 4-20, 2015. [14] Z.-C. Xu, C.-Q. Tang, and Z. Zhang, " Building topology-aware overlays using global soft-state,” in Proc. of the 23rd Intl. Conf. on Distributed Computing Systems, Providence, 2003, pp. 1-9. [15] Z. Li, G. Xie, K. Hwang, and Z. Li, " Churn-resilient protocol for massive data dissemination in P2P networks,” IEEE Trans. on Parallel and Distributed Systems, vol. 22, no. 8, pp. 1342-1349, 2011.