• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
An Enhanced SIP Proxy Server for Wireless VoIP in Wireless ...
 

An Enhanced SIP Proxy Server for Wireless VoIP in Wireless ...

on

  • 407 views

 

Statistics

Views

Total Views
407
Views on SlideShare
407
Embed Views
0

Actions

Likes
0
Downloads
1
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    An Enhanced SIP Proxy Server for Wireless VoIP in Wireless ... An Enhanced SIP Proxy Server for Wireless VoIP in Wireless ... Document Transcript

    • ADVANCES IN WIRELESS VOIP An Enhanced SIP Proxy Server for Wireless VoIP in Wireless Mesh Networks Bo Rong, International Institute of Telecommunications Yi Qian, National Institute of Standards and Technology Hsiao-Hwa Chen, National Sun Yat-Sen University ABSTRACT inherent combination of wireless infrastructure, user mobility, and heterogeneous network com- The wireless mesh network (WMN) has puting [1–3]. In this article, we mainly address emerged recently as a promising technology for the following three challenging issues. First, the next-generation wireless networking. In WMNs, interaction between a WMN and the IP core it is important to provide high quality multime- network can increase the signaling complexity dia service in a flexible and intelligent manner. and cause long, call set up delays. Second, the To address this issue in this article, we study the requirements of access bandwidth change from Session Initiation Protocol (SIP) for wireless time to time in WMNs due to the mobility of voice over IP (VoIP) applications. Especially, we users and the variation of wireless channel con- investigate the technical challenges in WMN ditions. Therefore, it is necessary to design a VoIP systems and propose a design of an dynamic access bandwidth prediction and reser- enhanced SIP proxy server to overcome them. vation scheme. Third, a call admission control An analysis of the signaling process and a study (CAC) mechanism should be implemented in of simulation results have shown the advantages case there is a distinction between the actual of our proposed approach. access bandwidth requirements and the predict- ed/reserved access bandwidth condition. INTRODUCTION To overcome these challenges, we further propose to build an enhanced SIP proxy server, In the last few years, the wireless mesh network which we describe in this article. The enhanced (WMN) has drawn significant attention in the SIP proxy server employs common open policy research community as a fast, easy, and inexpen- service (COPS) to dynamically reserve the access sive solution for broadband wireless access [1–3]. bandwidth in the IP core network for all SIP ter- However, there are still many challenges in the minals in a WMN. Moreover, the enhanced SIP design of WMNs. One of the most important proxy server contains two special modules to issues is how to efficiently support a wireless deal with traffic prediction and call admission voice over IP (VoIP) application, which is control problems. The rest of this article is orga- expected to be one of the killer applications for nized as follows. We first introduce the back- future wireless networks. ground of WMNs and SIP-based VoIP. We then Many researchers advocate Session Initiation discuss the challenges of deploying SIP in a Protocol (SIP) as a feasible signaling solution for WMN and develop an enhanced SIP proxy serv- VoIP applications, because it is simpler and er to deal with these challenges. Finally, we eval- more efficient than H.323 [4, 5]. For example, uate the performance of our proposed approach SIP has been selected as the call control proto- and conclude the article. col for third generation (3G) IP-based mobile networks [6]. In this article, we address how to SIP-BASED VOIP IN WMNS deploy SIP in WMNs to support quality of ser- vice (QoS), guaranteed multimedia communica- WIRELESS MESH NETWORKS tion. Particularly, we assume that a WMN is As shown in Fig. 1, a WMN consists of two types connected through a gateway to the IP core net- of nodes: mesh routers and mesh clients. The work, which employs multi-protocol label switch- mesh routers form the infrastructure of a mesh ing (MPLS) technology [7]. backbone for mesh clients. In general, mesh To deploy SIP in such a network architecture, routers have minimal mobility and operate just we must deal with many new technical chal- like a network of fixed routers, except that they lenges that have never been faced in wired net- are connected by wireless links through wireless works. These new challenges are raised by the technologies such as IEEE 802.11. We can see 108 0163-6804/08/$25.00 © 2008 IEEE IEEE Communications Magazine • January 2008
    • from Fig. 1 that the WMN can access the Inter- net through a gateway mesh router that is con- r0 r5 Wireless mesh nected to the IP core network with physical backbone r11 ri wires. In this study, we assume that the IP core network employs MPLS technology. MPLS oper- r10 r12 ates at an open systems interconnection (OSI) r1 r13 r4 r6 model layer that lies between traditional defini- tions of layer 2 (data link layer) and layer 3 (net- Coverage area of work layer). It was designed to provide a unified r2 r9 r14 mesh router ri data-carrying service for both circuit-based r7 clients and packet-switching clients. Many Gi: Wireless mesh r15 researchers recommended MPLS as a reliable r3 router with way to provide QoS guaranteed services [7]. gateway Gi r16 ri: Wireless mesh In WMNs, every mesh router is equipped r8 router with a traffic aggregation device (similar to an r17 Wireless mesh 802.11 access point) that interacts with individu- clients al mesh clients. The mesh router relays the aggregated data traffic of mesh clients to and MPLS-based IP core network from the IP core network. Typically, a mesh router has multiple wireless interfaces to com- municate with other mesh routers, and each wireless interface works, corresponding to one I Figure 1. An example of wireless mesh network. wireless channel. These channels have different characteristics due to the inherent features of a wireless environment. In practice, wireless inter- faces are usually running on different frequen- SIP proxy server SIP proxy server cies and built on either the same or different wireless access technologies, such as IEEE 802.11a/b/g/n. It is also possible that directional antennas are employed on some interfaces to Wireless mesh MPLS-based IP Wireless mesh establish wireless channels over long distance. network LER core network LER network SIP SIP terminal terminal SIP-BASED VOIP SIP is defined in RFC 2543 as an Internet Engi- neering Task Force (IETF) standard for multi- media conferencing over IP. It is an I Figure 2. The deployment of SIP based VoIP in WMNs. ASCII-based, application-layer control protocol that can be used to establish, maintain, and ter- minate calls between two or more end points. TECHNICAL CHALLENGES OF Just like other VoIP protocols, SIP is designed to provide the functions of signaling and session DEVELOPING SIP-BASED management within a packet telephony network. VOIP IN WMNS Signaling enables call information to be carried across network boundaries. Session management In this article, we address the issues of how to provides the ability to control the attributes of use SIP to support the wireless VoIP in WMN- an end-to-end call. accessed IP networks. According to the original Compared to H.323, SIP is a much more design, SIP sets up and tears down only sessions streamlined protocol, developed specifically for with a minimal focus on the management of IP telephony [5]. SIP is simpler and more effi- active sessions. To deploy SIP in WMNs, we cient than H.323, and it takes advantage of exist- must face many new challenging issues that are ing protocols to handle certain parts of the caused by the instability of the wireless environ- process. For example, media gateway control ment and by user mobility. In this article, we protocol (MGCP) is used by SIP to establish a mainly study three technical challenges in WMN gateway connecting to the public-switched tele- SIP deployment, that is, call set up delay, access phone network (PSTN) system. bandwidth prediction/reservation, and call admis- Figure 2 demonstrates the deployment of SIP- sion control. based VoIP in WMNs. Here, the SIP proxy server is an intermediate device that receives SIP CALL SET UP DELAY requests from a client and then forwards the In the real world, a WMN usually serves as an requests on behalf of the client. Basically, proxy access network to the Internet. To provide guar- servers receive SIP messages and forward them to anteed QoS, IP core networks are built with the next SIP server in the network. Proxy servers wired technologies, such as MPLS. Moreover, can provide functions such as routing, reliable most VoIP applications intend to go out of their request retransmission, authentication, authoriza- own local WMNs for counterparts in the Inter- tion, and security. Moreover, each WMN is con- net. Therefore, when SIP is used to set up a nected to the MPLS-based IP core network VoIP session, it must face a heterogeneous net- through a label edge router (LER) that operates work environment. This heterogeneous network at the edge of an MPLS network and uses routing environment increases the complexity of the sig- information to assign labels to datagrams and naling process and causes a long call set up then forwards them into the MPLS domain. delay. IEEE Communications Magazine • January 2008 109
    • cost scheme may not ensure that all users are satisfied, although it is able to reduce the expense of WMN operators. Call admission control Dynamic access bandwidth reservation requires the prediction of outgoing traffic load in WMNs. Because there always exists a distinc- tion between the exact access bandwidth require- SIP messages ment and the predicted access bandwidth requirement, the call admission control mecha- SIP and COPS Access bandwidth nism must be implemented. protocol stacks prediction CALL ADMISSION CONTROL COPS messages A CAC mechanism must be employed when the predicted and reserved access bandwidth is dif- ferent from the real one. CAC is used to accept Access or reject connection requests based on the QoS bandwidth requirements of these connections and the sys- negotiation tem state information. CAC prevents oversub- scription of VoIP networks and is a concept that applies only to real-time media traffic but not to I Figure 3. The framework of the enhanced SIP proxy server. data traffic. A CAC mechanism complements the capabil- ities of QoS tools to protect audio/video traffic Without losing the capability of generalizing, from the negative effects of other audio/video this article studies the scenario where a WMN is traffic and to keep excessive audio/video traffic connected to the MPLS-based IP core network. away from the network. CAC can also help wire- We assume that the MPLS network runs with less mesh networks to provide different types of traffic engineering capability, which is an essen- traffic load with different priorities by manipu- tial step to achieve high efficiency. In traffic lating their blocking probabilities. engineering-enabled MPLS networks, Con- straint-based Routing Label Distribution Proto- col (CR-LDP) or Resource Reservation Protocol AN ENHANCED SIP PROXY SERVER with Traffic Engineering Extensions (RSVP-TE) is employed to set up a label-switched path FOR WIRELESS VOIP (LSP) dynamically for a connection with QoS Conventionally, a proxy server is an optional SIP requirements. As a result, the total session set component that handles routing of SIP signaling up delay of a VoIP call should be the sum of SIP but does not initiate SIP messages. Proxy servers signaling and MPLS signaling times if one SIP also can provide some auxiliary functions such as client in a WMN wants to communicate with its authentication, authorization, reliable request counterpart in another WMN through MPLS- retransmission, and security. To overcome the based IP core network. technical challenges in wireless VoIP deploy- ment, we develop an enhanced SIP proxy server ACCESS BANDWIDTH PREDICTION AND with the framework shown in Fig. 3. RESERVATION Particularly, the enhanced SIP proxy server utilizes COPS messages to negotiate with the When designing a SIP architecture for WMNs, MPLS LER about the overall access bandwidth we must note two facts: requirement on behalf of all SIP terminals in a • The users in WMNs are free to move to WMN (not only one SIP terminal). Then, the anywhere at anytime. LER exchanges traffic engineering signaling with • The wireless channel conditions may vary other routers inside the MPLS core network to from time to time. set up the corresponding LSPs. In this way, the Clearly, these two facts can result in varying LSPs required by SIP telephony are set up in the access bandwidth requirements in WMNs. MPLS core network before SIP calls are made. To accommodate this variation, the best way As a result, the SIP call set up delay in the MPLS is to let WMN gateway mesh routers dynamically network is decreased significantly. reserve access bandwidth from the IP core net- We use a set of time marks {t0, t1, t2, …, tn–1, work, because the fixed bandwidth reservation tn, tn+1, …} to distinguish the time instances of approach is not efficient in this scenario. For the system. If the enhanced SIP proxy server example, there can be two straightforward ways knows that the overall access bandwidth require- to reserve the fixed access bandwidth for vari- ment of its WMN during [t n–1 , t n ] is exactly able requirements. The first way is called the Bn–1,n, then at time tn–1, the enhanced SIP proxy optimal user satisfaction scheme, which reserves server negotiates with the MPLS network to the maximum bandwidth that a WMN ever obtain Bn–1,n outgoing bandwidth by using COPS requires. The second way is called the optimal messages. As time goes by, if the enhanced SIP cost scheme, which reserves the minimum band- proxy server knows that the overall bandwidth width that a WMN ever requires. Nevertheless, requirement of the WMN during [t n , t n+1 ] both of these methods have their shortcomings. changes to Bn,n+1, then at time tn, the enhanced The optimal user satisfaction scheme is not eco- SIP proxy server should renegotiate with the nomic, although it can always provide enough MPLS network to increase/decrease the band- access bandwidth for WMN users. The optimal width requirement to Bn,n+1. 110 IEEE Communications Magazine • January 2008
    • Enhanced SIP Enhanced SIP A CAC mechanism proxy server proxy server complements the SIP SIP capabilities of QoS terminal terminal tools to protect MPLS-based IP audio/video traffic Local client core network Local client network network from the negative LER LER effects of other audio/video traffic If accepted If accepted and to keep INVITE INVITE INVITE INVITE INVITE excessive audio/video If declined If declined traffic away from Decline decline the network. Decline Decline 180 ringing 180 ringing 180 ringing 200 OK 200 ACK 200 ACK ACK ACK ACK Traffic stream If bandwidth negotiation is needed Cops REQ Cops REQ Cops DEC Cops DEC Cops REQ Cops REQ Cops DEC Cops DEC I Figure 4. The signaling flow with the enhanced SIP proxy server. However, it is impossible for the enhanced the literature, extensive studies have been con- SIP proxy server to know the exact value of ducted on access bandwidth prediction and call Bn–1,n before the time instance of tn–1. Usually, admission control. As a result, our proposed the enhanced SIP proxy server can employ only enhanced SIP proxy server can directly inherit a certain bandwidth prediction algorithm to give these research results. For access bandwidth pre- an approximate value of B n–1,n , which can be diction, a multiresolution finite-impulse-response ^ ^ defined as B n–1,n. If B n–1,n < Bn–1,n during [tn–1, (FIR) neural-network-based learning algorithm t n ], the WMN does not have enough outgoing was developed in [8], using the maximal overlap bandwidth to accommodate all SIP calls, and the discrete wavelet transform (MODWT). This enhanced SIP proxy server must utilize a call algorithm is a good choice for the enhanced SIP admission control mechanism to decline some of proxy server, because it has satisfactory trade-off ^ the call requests. In contrast, if B n–1,n > Bn–1,n} between prediction accuracy and computational during [tn–1, tn], some of the outgoing bandwidth complexity. On the other hand, the design of call resource of the WMN would be wasted. From admission control can be formulated as an opti- the above discussion, we can conclude that the mization problem, where the demands of both algorithms of access bandwidth prediction and the WMN service provider and the user are call admission control running on an enhanced taken into account. To solve this optimization SIP proxy server are critical to our approach. In problem, the authors of [9] proposed a utility- IEEE Communications Magazine • January 2008 111
    • decides whether this SIP call request is admitted. 5 If the call request is admitted, an enhanced SIP With traditional SIP proxy server proxy server will forward the original INVITE With enhanced SIP proxy server message to the callee; otherwise, it simply sends the caller a DECLINE message to drop the call. 4 Furthermore, whether the call is admitted or not, it is registered in the enhanced SIP proxy Average SIP call setup delay (s) server for the purposes of access bandwidth pre- 3 diction and call admission control in the future. If the access bandwidth must change, the enhanced SIP proxy server uses COPS messages 2 to negotiate with the MPLS-based IP core net- work to set up new LSPs with the required band- width. The COPS Protocol is part of the Internet protocol suite as defined by IETF RFC 2748. 1 COPS specifies a simple client/server model for supporting policy control over QoS signaling protocols (e.g., RSVP). As shown in Fig. 4, the 0 enhanced SIP proxy server uses COPS request 0 0.5 1 1.5 2 2.5 3 3.5 4 (REQ) and COPS decline (DEC) messages to Time (h) make bandwidth negotiations with the MPLS core network in an on-demand manner. I Figure 5. The average call setup delay of SIP-based wireless VoIP over four The previous discussions clearly show that the hours. approach of an enhanced SIP proxy server can reduce considerably the call set up delay, because the LSPs required by SIP telephony are set up in the MPLS-based IP core network 6 before SIP calls start. SIP signaling delay (traditional SIP proxy server) MPLS signaling delay (traditional SIP proxy server) SIP signaling delay (enhanced SIP proxy server) SIMULATION RESULTS 5 MPLS signaling delay (enhanced SIP proxy server) To further demonstrate the advantages of our Average SIP and MPLS signaling approach, we conducted a simulation study to 4 compare the performance of a traditional SIP proxy server and an enhanced SIP proxy server in terms of call set up delay. We used OPNET delay (s) 3 Modeler 11.0 to simulate the network environ- ment as shown in Fig. 1. In the simulation, we studied the case of medium traffic load, which is 2 incurred by real-time multimedia applications. We programmed the traditional SIP proxy server 1 according to [4] and the enhanced SIP proxy server according to the architecture proposed in this article. For simplicity, in our simulation, the 0 enhanced SIP proxy server employs a complete 0 0.5 1 1.5 2 2.5 3 3.5 4 sharing (CS) CAC policy, which means that an Time (h) incoming connection is accepted if sufficient bandwidth resources are available. I Figure 6. The decomposition of call setup delay. Figure 5 demonstrates the average call set up delay of SIP-based wireless VoIP during four hours. It is seen that the SIP call set up delay constrained, greedy approximation algorithm, varies between three and four seconds when a which can be easily implemented in the traditional SIP proxy server is employed. On the enhanced SIP proxy server. other hand, the delay is as low as 0.6 second or even less when an enhanced SIP proxy server is PERFORMANCE ANALYSIS employed. Figure 6 illustrates the decomposition of call SIGNALING PROCESS WITH AN set up delay, which includes the SIP signaling ENHANCED SIP PROXY SERVER delay and the MPLS signaling delay. As we can see, an enhanced SIP proxy server generates a Figure 4 shows the signaling flow in the pro- much shorter call set up delay than a traditional posed SIP architecture with an enhanced SIP SIP proxy server, because it has a significant proxy server. The call set up starts with a stan- decrement of MPLS signaling delay. dard SIP INVITE message sent from the caller It is noted that we also should be concerned to the local enhanced SIP proxy server in a about the performance of an enhanced SIP WMN. This message carries the callee URL in a proxy server in terms of access bandwidth pre- SIP header and the QoS requirements of a SIP diction and call admission control. From existing call-in body Session Description Protocol (SDP). studies as previously mentioned, an enhanced Regarding the caller ID, the QoS require- SIP proxy server can directly borrow an access ments, and the remaining outgoing bandwidth in bandwidth prediction algorithm and call admis- a local WMN, the enhanced SIP proxy server sion control algorithm. Therefore, the perfor- 112 IEEE Communications Magazine • January 2008
    • mance of access bandwidth prediction and call in the Department of Electrical Engineering, Ecole de tech- nologie superieure, Universite du Quebec for three years, admission control depends on which algorithm and then as a postdoctoral fellow in the Department of To further the enhanced SIP proxy server employs. Electrical and Computer Engineering, University of Puerto Rico at Mayaguez for one and a half years. demonstrate the advantages of our CONCLUSION Y I Q IAN [M’95, SM’07] (yqian@nist.gov) received a Ph.D. degree in electrical engineering with a concentration in approach, we In this article, we investigated the deployment of telecommunication networks from Clemson University. He SIP-based VoIP in WMNs. We first discussed is with the National Institute of Standards and Technology, conducted a the technical challenges in a wireless VoIP sys- in Gaithersburg, MD. His current research interests include network security, network design, network modeling, simu- simulation study to tem, such as call set up delay, access bandwidth lations and performance analysis for next generation wire- prediction and reservation, call admission con- compare the less networks, wireless sensor networks, broadband trol, and so on. We then proposed a novel satellite networks, optical networks, high-speed networks performance of a approach of an enhanced SIP proxy server to and the Internet. He has publications and patents in all these areas. He was an assistant professor in the Depart- traditional SIP proxy deal with these challenges. The analysis of sig- ment of Electrical and Computer Engineering, University of naling process and the study of simulation results Puerto Rico at Mayaguez (UPRM) between July 2003 and server and an have shown the advantages of our proposed July 2007. At UPRM, he taught courses on wireless net- enhanced SIP proxy approach. works, network design, network management, and net- work performance analysis. Prior to joining UPRM in July server in terms of call 2003, he worked for several start-up companies and con- REFERENCES sulting firms, in the areas of voice over IP, fiber optical set up delay. switching, Internet packet video, network optimizations, [1] I. F. Akyildiz and X. Wang, “A Survey on Wireless Mesh and network planning as a technical advisor and a senior Networks,” IEEE Commun. Mag., vol. 43, no. 9, Sept. consultant. He also worked several years for the Wireless 2005, pp. S23–S30. Systems Engineering Department, Nortel Networks in [2] A. Raniwala and T. Chiueh, “Architecture and Algo- Richardson, Texas as a senior member of the scientific staff rithms for an IEEE 802.11-based Multi-Channel Wireless and as a technical advisor. While at Nortel, he was a pro- Mesh Network,” IEEE INFOCOM 2005, vol. 3, Mar. ject leader for various wireless and satellite network prod- 2005, pp. 2223–34. uct design projects, customer consulting projects, and [3] H. Jiang et al., “Differentiated Services for Wireless advanced technology research projects. He was also in Mesh Backbone,” IEEE Commun. Mag., vol. 44, no. 7, charge of wireless standard development and evaluations. July 2006, pp. 113–19. He is the author of the book Information Assurance — [4] J. Rosenberg et al., “SIP: Session Initiation Protocol,” Dependability and Security in Networked Systems (Morgan RFC 3261 IETF, June 2002. Kaufmann, 2007). He is a member of ACM. [5] U. Black, Voice Over IP, Prentice Hall, 2000. [6] 3GPP, “Technical Specification Group Services and Sys- HSIAO-HWA CHEN [S’89, M’91, SM’01] (hshwchen@ieee.org) tem Aspects; Network Architecture (Release 5),” Techni- received B.Sc. and M.Sc. degrees from Zhejiang University, cal Report TS23.002, 3GPP, Mar. 2002. China, and a Ph.D. degree from the University of Oulu, Fin- [7] T. Li, “MPLS and the Evolving Internet Architecture,” land in 1982, 1985, and 1990, respectively, all in electrical IEEE Commun. Mag., vol. 37, no. 12, Dec. 1999, pp. engineering. He is currently a full professor and was the 38–41. founding director of the Institute of Communications Engi- [8] V. Alarcon-Aquino and J. A. Barria, “Multiresolution FIR neering at the National Sun Yat-Sen University, Taiwan. He Neural-Network-Based Learning Algorithm Applied to has authored or co-authored over 200 technical papers in Network Traffic Prediction,” IEEE Trans. Systems, Man major international journals and conferences, five books, and Cybernetics, Part C: Applications and Reviews, vol. and several book chapters in the areas of communications, 36, no. 2, Mar. 2006, pp. 208–20. including the books, [[Next Generation Wireless Systems [9] B. Rong, Y. Qian, and K. Lu, “Integrated Downlink and Networks]] and [[The Next Generation CDMA Tech- Resource Management for Multiservice WiMAX Net- nologies]], both published by John Wiley and Sons in 2005 works,” IEEE Trans. Mobile Computing, vol. 6, no. 6, and 2007, respectively. He has been an active volunteer for June 2007, pp. 621–32. various IEEE technical activities for over 20 years. Currently, he is serving as the chair of IEEE Communications Society BIOGRAPHIES Radio Communications Committee. He served or is serving as symposium chair/co-chair of many major IEEE confer- BO RONG [M’07] (bo.rong@ieee.org) received a B.S. degree ences, including VTC, ICC, GLOBECOM, WCNC, and so on. from Shandong University in 1993, an M.S. degree from He served or is serving as associate editor and/or guest edi- Beijing University of Aeronautics and Astronautics in 1997, tor of numerous important technical journals in communi- and a Ph.D. degree from Beijing University of Posts and cations. He is serving as the chief editor (Asia and Pacific) Telecommunications in 2001. Currently, he is a researcher for Wiley’s Wireless Communications and Mobile Comput- at the International Institute of Telecommunications, Mon- ing (WCMC) Journal and Wiley’s International Journal of treal, Canada. His current research interests focus on mod- Communication Systems. He is the Editor-in-Chief of Wiley eling, simulation, and performance analysis for Security and Communication Networks journal (www.inter- next-generation wireless networks. After receiving his science.wiley.com/journal/security). He is also an adjunct Ph.D., he worked as a software engineer for a start-up professor at Zhejiang University, China and Shanghai Jiao company in Beijing for one year, as a postdoctoral fellow Tung University, China. IEEE Communications Magazine • January 2008 113