UMTS差別性服務等候機制效能應用二種佇列空間配置之研究

全文

(1)國立交通大學資訊管理研究所博士論文 UMTS 差別性服務等候機制效能應用二種佇列空間配置之研究 A Study on Performance Discussion of a UMTS Priority-based Queuing Scheme with two Queuing Buffer Allocations. 研究生：劉芳萍指導教授：楊. 千博士. 中華民國九十九年七月.

(2) UMTS 差別性服務等候機制效能應用二種佇列空間配置之研究. 研究生：劉芳萍. Student：Fanpyn Liu. 指導教授：楊千. Advisor：Chyan Yang. 國立交通大學資訊管理研究所博士論文. A Dissertation Submitted to Institute of Information Management College of Management National Chiao Tung University in partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Information Management July 2010 Hsinchu, Taiwan, the Republic of China. 中華民國九十九年七月. i.

(3) UMTS 差別性服務等候機制效能應用二種佇列空間配置之研究學生：劉芳萍. 指導教授：楊千. 國立交通大學資訊管理研究所. 摘要自美國、日本和歐洲各國提出第三代無線通訊系統協定標準，如 TD-SCDMA、CDMA2000 與 WCDMA / UMTS 以來，較廣為業界所看好的技術規格為 WCDMA / UMTS，因為目前使用人口最多的 GSM 歐洲系統業者也加入參與制定，且 GSM 是世界上最大的第二代移動通信網路，UMTS 勢必成為第三代行動通訊主流規格。3GPP 所提出 UMTS 的 3G 標準，可以確保 GSM 核心網路無縫 (Seamless) 接取到 3G 的核心網路，這種相容於過去 GSM / GPRS 系統的考量，對於系統業者不僅可投資延續價值，也兼顧商業上的經濟效益。 3G R5 以後的規格及 4G 網路已朝向全 IP 化的方向發展，並以 IP 作為主要通訊協定，因此，UMTS 的語音會話、影音串流、網頁互動及檔案傳輸的四種資料類型，都有其特定的屬性質，並賦予不同的封包傳輸順序，以達到符合各類型資料的服務等級 (Quality of Service, QoS) 的要求及目標，由於 IP 協定中的盡力而為服務(Best Effort Service)，對於嚴格要求低延遲、低抖動(Jitter)等服務的多媒體類或語音的資料型態，已無法滿足需求，因此，提供優化的封包轉送能力(Packet Forwarding)以達到服務等級要求，成為 UMTS 核心網路中一項重要的工作。本論文提出的方法是在 UMTS 系統核心網路中，對不同的資料類型，提供差異性服務及優先順序的封包傳送能力，利用兩種佇列緩衝記憶體空間配置機制：動態佇列緩衝空間配置(DQB)和邏輯佇列溢流緩衝空間配置(OQB)，另外，對各類型封包允入與允出佇列時的 ii.

(4) 機制，分別以優先順序為基礎方式(Priority-Based Queuing)允入佇列及連續性加權輪詢(Sequential-Weighted Round Robin)的允出佇列機制處理。透過 NS2 網路模擬軟體，對不同方案想定、封包大小及參數值進行模擬評估，根據模擬的結果，每一種類型服務的封包在效能(Throughput)、封包丟失率(drop probability)、延遲及抖動率與 IETF 所提出之差異性服務(Differentiated Service, DiffServ)比較，兩者可達到近似的效能。. 關鍵詞 UMTS、服務等級、優先佇列機制、邏輯佇列緩衝記憶體溢位、動態佇列緩衝記憶體配置、差異性服務、連續性加權輪詢佇列機制。. iii.

(5) A Study on Performance Discussion of a UMTS Priority-based Queuing Scheme with two Queuing Buffer Allocations. Student：Fanpyn Liu. Advisors：Dr. Chyan Yang. Institute of Information Management National Chiao Tung University. ABSTRACT As UMTS systems will evolve into an all-IP stage in the future, packet switching becomes a prerequisite for all UMTS applications. The 3GPP defines four types of UMTS traffic: conversational, streaming, interactive and background; each type of UMTS traffic has its QoS features. Differentiated Service (DiffServ) is QoS architecture for IP networks, their based on packet marking that allows packets to be prioritized according to Service Level Agreement (SLA) management. The DiffServ is a scalable architecture and is proposed to provide QoS guarantee services and scheduling packets forwarding for each class within the IP networks. According to four types of UMTS traffic, this study proposes a priority-based queuing scheme with two queuing buffer allocations, the dynamic queuing buffer (DQB) allocation and the overflow queuing buffer (OQB) allocation, to support packet forwarding of four types of UMTS traffic in a DiffServ method. In the proposed queuing scheme, two major modules, a priority-based enqueuing module and a sequential weighted round robin (SWRR) dequeuing module, base on the DQB iv.

(6) allocation and the OQB allocation to perform differentiated packet enqueuing / dequeuing jobs among four types of UMTS traffic. In this study, we use the ns2 (Network Simulator version 2) as the simulation platform to simulate several scenarios. Discussing the simulation results and analysis, we can find the proposed queuing scheme can base on packet transmission priorities of four types of UMTS traffic to support a differentiated packet forwarding behavior among UMTS traffic in a UMTS core network and the performance of UMTS traffic with a high priority always gets a better packet forwarding performance than that of UMTS traffic with a low priority. And, the differentiated packet forwarding behaviors with the proposed queueing scheme are similar to the packet forwarding behavior with the IETF DiffServ scheme.. Keyword: UMTS, priority-based queuing scheme, DQB allocation, OQB allocation, differentiated service packet forwarding, sequential weighted round robin queuing scheme.. v.

(7) 誌謝要開始寫「誌謝」這段，代表博士班的學生生涯要告一段落了，沒有胸懷壯志，也沒有驪歌高唱，只有一路走來受到師長照顧和幫忙，感激的心情無以言表，想說的感謝話有很多很多，想表達的誌謝辭也很長很長，遠遠超過這本論文的長度，非三言兩語可以道盡的。能完成這個學位，首先要感謝的人，是我的指導教授楊千博士，因為教授有教無類，願意收我為徒成為大師的門下，才得以一窺學術領域的深奧、作研究的嚴謹和人生智慧的哲理，生活中教授對於嚴肅的主題，常以詼諧幽默和逗趣的方式闡述其中道理，並也常不經意的在四兩撥千金的談笑間，剖析事情的角度分析其中的智慧，令我們佩服、震撼且感動不已。另外，我還要特別感謝的人是傅振華學長，因為他孜孜不倦、耐心和包容的指導，論文寫作、投稿甚至程式的撰寫、除錯和修改…等，才使得研究成果得以展現，完成這本論文的撰寫，心中有太多的感謝想要表達，千言萬語亦不足以道盡，能敘述得出來的也不及千萬分之一。徐道義教授在我就學期間，仍無時無刻關心我的研究狀況，並指導我作研究的技巧與態度，此外，也要感謝口試委員–羅濟群教授、劉敦仁教授和吳宗禮教授們逐字斧正論文內容，並犧牲週末假期參加學生計畫書及論文口試，並於論文修改期間不吝地提供我許多寶貴意見，啟發學生對問題的深究獲益良多，使本篇論文才能更加充實完善。另外，實驗室的學弟妹們–耿杰、建全、良駒、憲明、阿國、秋皓、士原和志棋…等一起通宵達旦作研究、討論、吃宵夜、打球、閒聊八卦及搞笑…等的革命情感，還有同班同學姐妹淘–翠娟、秀怡和明賢，以及俊龍和文茂平大哥在心情沮喪研究低潮時，相互勉勵加油打氣一起出遊、唱卡啦 OK 瘋狂的同窗之誼，這些都是我千年修來的福分，才能和他們成為同學；另外在我職場中的長官、同事也都是背後推動我前進堅持下去的動力。. vi.

(8) 最後，謹以此文獻給我摯愛的雙親、小弟和懷念的大弟博熙及愛犬(胖胖柴)，由於有他們默默的支持和陪伴，讓我能撐過這些煎熬的日子，年邁的父親期盼和我分享這份榮耀的日子已經很久了，感謝他至始至終對我的寬容和信心，並也要感謝我的愛犬們(蜜蜜和 PP)，因為牠們總是很有耐心靜靜的，陪我渡過一天又一天熬夜 K 書、寫論文的日子。要感謝的人及事有太多族繁不及備載，一路走來受到太多人的扶持和幫助，在博士生涯的研究歲月裏，也僅知學問與研究之不易，如訪隱者之不遇，時時喟嘆「只在此山中，雲深不知處」，知識的累積尚欠火候，回首這些年來，雖然其中沒有什麼值得特別炫耀的成果，但對我而言，是寶貴的人生歷練，是無數教誨、關愛和幫助的結果，本論文的完成絕非終點，文中的不足和淺顯之處則是我新的征途上，另一個負笈研究的新起點，感謝大家。. vii.

(9) Contents 摘要…… ..................................................................................................... ii ABSTRACT ................................................................................................ iv 誌謝…… .................................................................................................... vi Contents… ................................................................................................ viii List of Tables ............................................................................................... x List of Figures ............................................................................................. xi Abbreviations............................................................................................. xii Symbol Description ................................................................................... xv Chapter 1. Introduction .............................................................................. 1 1.1. Research Background and Motivation ................................................... 1 1.2. Related Research Work .......................................................................... 2 1.3. Organization of the Dissertation ............................................................ 4 Chapter 2. Literature Review .................................................................... 5 2.1. UMTS Services and Applications .......................................................... 5 2.2. Requirements for QoS ............................................................................ 6 2.2.1. QoS Mechanisms ......................................................................... 7 2.2.2. UMTS QoS Architecture.............................................................. 9 2.2.3. QoS Classes of UMTS applications ........................................... 10 2.3. UMTS with Differentiated Services (DiffServ)................................... 12 2.3.1. Integrated Services (IntServ) ..................................................... 13 2.3.2. Differentiated Services (DiffServ) ............................................. 14 Chapter 3. Review of Queue Management Mechanisms....................... 16 3.1 Passive Queue Management (PQM) ..................................................... 16 3.2 Active Queue Management (AQM)...................................................... 17 3.3. Random Early Detection (RED) .......................................................... 18 3.4. Priority-based Queueing ...................................................................... 19 3.5. Weighted Round Robin (WRR) Queueing .......................................... 20 Chapter 4. An Approach to Priority-based Queueing Scheme with Two Queueing Buffer Allocations ......................................... 22 viii.

(10) 4.1. Queueing Buffer Allocations in Priority-based Queueing Scheme ..... 22 4.1.1. A Dynamic Queueing Buffer (DQB) Allocation ....................... 23 4.1.2. An Overflow Queueing Buffer (OQB) Allocation .................... 24 4.2. A Priority-Based Enqueueing Module ................................................. 25 4.2.1. A Priority-based Packet Enqueueing Module with a DQB Allocation ................................................................................... 27 4.2.2. A Priority-based Packet Enqueueing Module with an OQB Allocation ................................................................................... 30 4.3. A WRR Packet Dequeuing Module ..................................................... 33 Chapter 5. Simulation and Results Analysis .......................................... 38 5.1. Simulation Topology and Parameters .................................................. 38 5.2. Simulation Results and Analysis ......................................................... 40 5.2.1. UMTS Packet Transmission in a Continuous Traffic Pattern ... 40 5.2.2. UMTS Packet Transmission in an Intermittent Traffic Pattern . 47 5.2.3. Summary .................................................................................... 52 Chapter 6. Conclusions ............................................................................. 54 References .................................................................................................. 57 Appendix.. .................................................................................................. 62 Appendix A: The DQB Source Code ......................................................... 63 Appendix B: The DQB Header File............................................................ 72 Appendix C: The OQB Source code ........................................................... 73 Appendix D: The OQB Header File ........................................................... 87 Appendix E: Simulation scenarios for DQB in a continuous traffic pattern ................................................................................... 89 Appendix F: Simulation scenarios for OQB in a continuous traffic pattern ................................................................................... 92 Appendix G: Simulation scenarios for DQB in a intermittent traffic pattern ................................................................................... 96 Appendix H: Simulation scenarios for OQB in a intermittent traffic pattern ................................................................................. 101 Responses to commitee comments ......................................................... 106. ix.

(11) List of Tables Table 1. Applications Enabling 3G Service Categories ............................... 6 Table 2. UMTS QoS features ...................................................................... 11 Table 3. Packet dequeuing turns received by all logical queuing buffers with the DQB / OQB allocations in a packet dequeuing cycle .. 34 Table 4. A summary of simulation parameters ........................................... 39 Table 5. A packet enqueueing / dequeueing statistic of UMTS traffic in a continuous transmission pattern ................................................. 41 Table 6. An average packet delay statistic of UMTS traffic in a continuous transmission pattern ................................................. 43 Table 7. An average packet jitter statistic of UMTS traffic in a continuous transmission pattern ................................................. 43 Table 8. A statistic of disorder packet transmission among UMTS traffic with the OQB allocation in a continuous traffic pattern............. 45 Table 9. A ranking weight statistic of UMTS application with both of the DQB allocation and the OQB allocation in a continuous traffic pattern ......................................................................................... 46 Table 10. A packet enqueuing / dequeuing statistic of UMTS traffic in an intermittent transmission pattern ................................................ 48 Table 11. An average packet delay statistic of UMTS traffic in an intermittent transmission pattern ................................................ 50 Table 12. An average packet jitter statistic of UMTS traffic in an intermittent transmission pattern ................................................ 50 Table 13. A statistic of disorder packet transmission among UMTS traffic with the OQB allocation in an intermittent traffic pattern ......................................................................................... 50 Table 14. A ranking weight statistic of UMTS application with both of the DQB/OQB allocation in an intermittent traffic pattern ........ 51. x.

(12) List of Figures Figure 1. IP QoS mechanisms ....................................................................... 7 Figure 3. The QoS functions and can be implemented on a router .............. 8 Figure 2. Diagram of the infrastructure of a UMTS network ....................... 8 Figure 4. UMTS QoS Architecture ............................................................... 9 Figure 5. Sketch map of Best effort and IntServ ........................................ 14 Figure 6. DiffServ Control Architecture ..................................................... 15 Figure 7. Priority-based Queueing .............................................................. 20 Figure 8. First In First Out Queueing.......................................................... 21 Figure 9. Weighted Round Robin Queueing............................................... 21 Figure 10. A diagram of queuing buffer with the DQB allocation ............. 24 Figure 11. A diagram of queuing buffer with the OQB allocation ............. 25 Figure 12. A pseudo code of a two stage priority-based enqueuing process......................................................................................... 26 Figure 13. A diagram of arrival packets enqueueing process with the DQB allocation ........................................................................... 28 Figure 14. A pseudo code of the RED-based packet enqueuing process with the DQB allocation ............................................................. 29 Figure 15. A diagram of arrival packets enqueueing process with the OQB allocation ........................................................................... 31 Figure 16. The pseudo code of a RED-based packet enqueueing process with the OQB allocation ............................................................. 32 Figure 17. A diagram of the SWRR dequeuing module with the DQB allocation ..................................................................................... 33 Figure 18. A diagram of the SWRR dequeuing module with the OQB allocation ..................................................................................... 34 Figure 19. Flowchart of dequeueing turn transfer PDTT procedure .......... 36 Figure 20. Flowchart of the WRR packet dequeueing module .................. 37 Figure 21. A diagram of the simulation topology ....................................... 39 Figure 22. A statistics of packet dequeueing volume in a continuous transmission pattern .................................................................... 42 Figure 23. A statistics of packet dequeuing rate in an intermittent transmission pattern .................................................................... 49 xi.

(13) Abbreviations 3G. 3rd Generation. 4G. 4th Generation. 3GPP. 3rd Generation Partnership Project. APD. Average Packet Delay. APJ. Average Packet Jitter. AQM. Active Queue Management. BAC. Background traffic. BS. Base Station. CDMA. Code Division Multiple Access. CN. Core Network. CON. Conversational traffic. DiffServ. Differentiated Services. DQB. Dynamic Queueing Buffer. DP. Dequeued Packets. DPR. Dequeued Packet Ratio. DPT. Disorder Packet Transmission. DRP. Dropped Packets. DRPR. Dropped Packet Ratio. ECN. Explicit Congestion Notification. EDGE. Enhanced Data Rates for GSM Evolution. EI. Evaluation Item. FDD. Frequency Division Duplex. FIFO. First In First Out. FTP. File Transfer Protocol. GBS. Guaranteed Buffer Size. GGSN. Gateway GPRS Support Node. GoS. Grade of Service. GPRS. General Packet Radio Service. GSM. Global System for Mobile communication xii.

(14) IETF. the Internet Engineering Task Force. INT. Interactive traffic. IntServ. Integrated Service. IP. Internet Protocol. IPTV. Internet Protocol Television. MT. Mobile Termination. NS2. Network Simulator 2. OQB. Overflow Queueing Buffer. PDTT. Packet Dequeueing Turn Transfer. PEP. Packet Enqueueing Probability. PHB. Per-Hop-Behavior. PKT. Packet. PQM. Passive Queue Management. PRI. Priority Queuing. PS. Packet Size. QoS. Quality of Service. RED. Random Early Detection. RFC. Request for Comment. RNC. Radio Network Controller. RR. Round Robin. SAP. Service Access Point. SDU. Service Data Unit. SLA. Service Level Agreement. SMS. Short Messaging Service. STR. Streaming traffic. SWRR. Sequential-WRR (Weighted Round Robin). TCL. Tool Command Language. TCP. Transport Control Protocol. TDD. Time Division Duplex. TE. Terminal Equipment Time Division Synchronous Code Division Multiple Access. TD-SCDMA. xiii.

(15) TPFPR. Total Packet Forwarding Performance Ranking. TT. Transmission Time. UDP. User Datagram Protocol. UE. User Equipment. UMTS. Universal Mobile Telecommunications System. UT. UMTS Traffic. UTRA. UMTS Terrestrial Radio Access. UTRAN. UMTS Terrestrial Radio Access Network. VoIP. Voice over Internet Protocol. WCDMA. Wireless Code Division Multiple Access. WRR. Weighted Round Robin. WS. Weight Sum. xiv.

(16) Symbol Description B. Bytes. D(i, j). The jitter between any two particular packets.. i or j. Send and receive times between two packets( i and j). ms. 10-3 seconds. Ri. The arrival time of packet i. Rj. The arrival time of packet j. Si. The sending time of packet i. Sj. The sending time of packet j. µs. 10-6 seconds. xv.

(17) Chapter 1. Introduction Nowadays, mobile communications have become popular communication fashions worldwide and are available to all. The evolution technologies over the last two decades has enabled the development of the ubiquitous mobile communication service, which can provide the mobile user with voice, data and multimedia services at any time, any place, and in any format. Hence, many applications and bandwidth requirements are proposed for mobile communications. The third generation (3G) and the proposed fourth generation (4G) mobile communication support a broadband mobile communication environment and diversified services.. 1.1. Research Background and Motivation The Universal Mobile Telecommunications System (UMTS), one of the 3G mobile communication standards developed in Europe, supports diversified mobile communication applications [1, 2]. According to 3GPP planning, an all IP-based architecture will eventually be adopted in the UMTS core network to support diversified 3G services [3]. For reducing the costs to increase the revenue, the network service providers plan to merge the data communication network and telephony communication network and develop the all-IP networks. In an all-IP network, traffic is packetized and transmitted within a UMTS core network and external IP networks [4-7]. These traffic types of UMTS applications can be divided into four classes: conversational, streaming, interactive, and background. For 3G applications, the 3GPP defines four types of traffic, each with different Quality of Service (QoS) features. Differentiated services (DiffServ) must be supported within a UMTS core network to satisfy the required QoS of UMTS applications [8, 9]. The user will be judged on the basis of the these applications, on the other hand, network bandwidth and application priority will determine 1.

(18) the performance of the network under its control in terms of the cost of operating the network to support the agreed upon QoS requirements for its users. Today, people expect their mobile communication service to be available all the time with no deterioration in the service quality. Hence, it is imperative that the evaluation of the performance of the mobile network takes into account the survivability requirements. With a QoS solution based on different QoS classes the use of the mobile network resources can be optimized. The users of the new networks services are only interested in end-to-end QoS [10]. End-to-end services typically involve communication through external networks, which make it obligatory to be able to map UMTS QoS parameters to external network QoS parameters and vice versa. The main goal of this study is to analyze and compare the performance of different queuing scheme over UMTS core network gateway, using two queueing buffer allocations. We consider four classes of packets which have to be served, where every packet of class has priority forwarding behavior corresponding to the QoS requirement. The simulation results are collected and analyzed to understand the performance of the proposed queuing schemes. The priority-based queuing scheme is implemented using an ns2 simulator, through which several scenarios are assumed and simulated. Finally, a conclusion is provided.. 1.2. Related Research Work In the competitive communication world, network bandwidth is a precious and limited resource. As with any mobile communication network service, guaranteed service levels and network performance are critical factors. We investigate the queue scheduling scheme for the provision of QoS over UMTS core network gateway by network 2.

(19) performance. In fact network performance should not be measured in absolute quantities like dropped packets but the degree to which the network satisfies the service requirements of each application. Critical network parameters and performance to be used in specifying and assessing the speed, accuracy, dependability, and availability of IP pack transfer [11]. These parameters include packet drop, transfer delay, packet jitter, and throughput. In particular, the network performance and reliability are embodied by these QoS parameters. In order to reach the goal, the access and delivery rules have to be formalized and be able to make meaningful promises in end-to-end QoS. To realize satisfying QoS results, investigating the process of UMTS performance in a form of network simulation is a required in implementing a particular queuing discipline. The simulation can evaluate parameters associated with UMTS network performance; delay, packet loss, packet jitter and throughput in accessing new service and architecture. Packet loss information can be useful in tracking persistent congestion problem. The statistical characteristics of the lost packets are based on established loss models such as Gilbert, Poisson, and Bernoulli [12, 13]. The probability of packet loss resulted from each of sent packets. That is: Packet Loss Ratio=. Number _ of _ Lost _ Packets Total _ Transmitted _ Packets. ……….. (1). Jitter is defined as the mean deviation of the difference in packet spacing at the receiver compared to the packet spacing at the sender for a pair of packets. This value is equivalent to the deviation in transit time for a pair of packets. The jitter metric is defined as the difference in send and receive times between two packets, i and j. The difference, D(i, j), provides the jitter between any two particular packets, however, a jitter value which measures the accumulated jitter over all packets is required.. 3.

(20) D (i, j) = (Rj - Sj) - (Ri - Si). ……….. (2). where Si=the sending time of packet i; Ri=the arrival time of packet i; Sj=the sending time of packet j; Rj=the arrival time of packet j and. j ＞i. The jitter of the response time is very important for real-time applications such as telephony. Web browsing and mail are fairly resistant to jitter, but any kind of streaming media (voice, video, music) is quite susceptible to jitter. Jitter is a symptom that there is congestion, or not enough bandwidth to handle the traffic.. 1.3. Organization of the Dissertation This dissertation is organized as follows. Related research works and literature review are surveyed in Chapter 2. Chapter 3 introduces some queue management mechanisms related to construct a queueing scheme. In Chapter 4 and Chapter 5, they are the most important core of this dissertation. Here we describe our queueing module and priority based queueing scheme with two different queueing buffer allocations in a UMTS core network gateway. Then, chapter 5 presents some simulation results to compare performance of priority-based queueing scheme with two queueing buffer allocation. Finally, we conclude our research and summaries in Chapter 6.. 4.

(21) Chapter 2. Literature Review 2.1. UMTS Services and Applications UMTS is a 3G broadband, packet-based transmission of text, digitized voice, video, video conferencing, IPTV (Internet Protocol television) and multimedia at data rates up to 2Mbps, can be reached. Higher bit rates naturally facilitate some new services, such as video telephony and quick downloading of data [14, 15]. If there is to be a killer application, it is most likely to be quick access to information and its filtering appropriate to the location of a user. Often the requested information is on the Internet, which calls for effective handing of TCP/UDP/IP traffic in the UMTS network. At the start of the UMTS era almost all traffic will be voice, but later the share of data will increase. It is, however, difficult to predict the pace at which the share of data will start to dominate the overall traffic volume. At the same time that transition from voice to data occurs, traffic will move from circuit-switched connections to packet-switched connections. At the start of UMTS service not all of the QoS functions will be implemented and therefore delay-critical applications such as speech and video telephony will be carried on circuit-switched bearers. Later, it will be possible to support delay-critical services as packet data with QoS functions [16]. Compared to GSM and other existing mobile networks, UMTS provides a new and important feature, namely it allows negotiation of the properties of a radio bearer. Attributes that define the characteristics of the transfer may include throughput, transfer delay and data error rate. To be a successful system, UMTS has to support a wide range of applications that possess different QoS requirements. At present it is not possible to predict the nature and usage of many of these applications. Therefore it is neither possible nor sensible and optimise UMTS to only one set of applications. Table 1 illustrates how the applications enabling 3G service categories [17]. 5.

(22) Table 1. Applications Enabling 3G Service Categories Service Edutainment Telematics Location B2C Office TeleCategories and Telemetry Based Service Extension medicine Application Infotainment Monitoring H H H H M L Multimedia H H H L L M m-commerce Unified L L L H M L messaging M M M H H L VoIP Interactive H H M L L L Broadcasting M H H H M H IP Access H H L M L H Positioning Legends: L: Low importance, M: Medium importance, H: High importance. 2.2. Requirements for QoS 3GPP has specified high level requirements for UMTS QoS. These requirements are divided into three categories – end user, general and technical requirements. According to the 3GPP planning, an all IP-based architecture will be adopted in the UMTS core network eventually to support diversified 3G services. In an all-IP network, traffic is packetized and transmitted within a UMTS core network and external IP networks [3]. For 3G applications, the 3GPP defines four types of traffic; their QoS features are different. DiffServs must be supported within a UMTS core network to satisfy the required QoS of UMTS applications [18]. The performance of packet forwarding process is one of the important factors that affect QoS of UMTS applications. For UMTS packet forwarding process in a UMTS core network gateway, a queuing scheme always plays an important role. A proper priority-based queuing scheme within a UMTS core network can support packet forwarding in a DiffServ way for UMTS applications. This study focuses on performance comparison of priority-based queuing scheme with two different queuing buffer allocations in a 6.

(23) UMTS core network gateway. The proposed queuing scheme is implemented in an ns2 simulator and several scenarios are assumed and simulated. The simulation results are collected and analyzed to observe the performance of the proposed queuing scheme with different queuing buffer allocations. Finally, a conclusion is reached.. 2.2.1. QoS Mechanisms QoS is one of the most important issues in networks in general, and particularly so in the Internet and other IP networks. QoS deals with the strict management of traffic such that guarantees can be made and SLAs (Service Level Agreement) between customers and service providers can be observed. In the case of packet switching, QoS basically guarantees that a packet will travel successfully between any two points. In packet-switched networks, these parameters need to be controlled in order to guarantee QoS, including latency end to end, jitter, loss, sequencing (i.e., the order of delivery of the packets), and errors etc.. Figure 1 shows the four main IP QoS mechanisms: classification (used for packet identification), conditioning (used for traffic shaping), queue management (used to manage the queue depth), and queue scheduling (used for packet scheduling) [19].. Figure 1. IP QoS mechanisms. 7.

(24) When implementing QoS, a common mechanism used is queuing, We survey a variety of queuing strategies that manage resources where congestion might occur. In UMTS networks, the transition from a TE (Terminal Equipment, such as mobile phone, Laptop…etc) via RNC (Radio Network Controller) to a UMTS core network makes the gateway in between a congestion point.. Figure 2. Diagram of the infrastructure of a UMTS network. Figure 3. The QoS functions and can be implemented on a router. In such cases, queuing might be configured on the gateway at the network edge. Figure 2 shows diagram of the infrastructure of a UMTS network [2]. Figure 3 illustrates the QoS mechanisms can be implemented the UMTS core network and briefly describes the QoS functions [19]. 8.

(25) 2.2.2. UMTS QoS Architecture Network services are considered end-to-end, this means from a Terminal Equipment (TE) to another TE. An End-to-End service may have a certain QoS which is provided for the user of a network service. It is the user that decides whether he is satisfied with the provided QoS or not. To realize a certain network QoS a Bearer Service with clearly defined characteristics and functionality is to be set up from the source to the destination of a service. A bearer service includes all aspects to enable the provision of a contracted QoS. These aspects are among others the control signaling, user plane transport and QoS management functionality. A UMTS bearer service layered architecture is depicted in Figure 4, each bearer service on a specific layer offers its individual services using services provides by the layers below [10].. Figure 4. UMTS QoS Architecture Legends: TE: Terminal Equipment, MT: Mobile Termination, CN: Core Network, UTRAN: UMTS Terrestrial Radio Access Network, EDGE: Enhanced Data Rates for GSM Evolution, lu: the lu interface to connect UTRAN and CN UTRA: Universal Terrestrial Radio Access FDD: Frequency Division Duplex TDD: Time Division Duplex 9.

(26) 2.2.3. QoS Classes of UMTS applications According to UMTS applications’ features, four QoS types of traffic, conversational, streaming, interactive, and background, are defined by the 3GPP. Each type of traffic has its QoS features. Four types of UMTS traffic are described as follows [10]. Conversational Traffic VoIP, videoconference, and video telephony are the typical conversational traffic. Real time is the most important characteristic for conversational traffic. Moreover, to support conversational traffic, a low delay and low jitter service is required. The transfer delay will be significantly lower than the round trip delay of an interactive application. The acceptable packet transfer delay for conversational traffic is very stringent. Streaming traffic Watching a real time video or listening to a real time audio from a video/audio server through the UMTS network is a typical streaming application. Generally speaking, streaming traffic is one way packet transport, and packets are transmitted to users in real time. Streaming traffic require a small delay variation. However, low transfer delay is not required. Interactive traffic When an end user accesses information or data from a server through a UMTS network, it belongs to interactive traffic. Typical interactive UMTS applications include web browsing, chatting room, ICQ, and telnet; they require low packet loss rate. A reasonable packet transfer delay is allowed for interactive traffic. In general, interactive UMTS applications are classical data communication applications that are characterized by the request/response pattern of the end-user. Round trip delay time is one of the key attributes. A low bit error rate of 10.

(27) transmitted packets is another characteristic. Background traffic Background delivery of E-mails, SMS, FTP, and reception of measurement records within a UMTS network are the typical background traffics. Background traffic transmission is also one of the classical data communication schemes that do not expect the data to be reached within a certain period of time. The QoS of background traffic requires a low packet loss rate and relaxed delay requirements, similar to the transmission of best effort traffic. Therefore, packet transfer of background traffic is more or less time insensitive with a low bit error rate. Table 2. UMTS QoS features Traffic class. Fundamental features. - Preserve time relation between information entitles of the stream Conversational -Sensitive to packet delay -Conversational pattern - Preserve time relation between information entities of the stream Streaming - Sensitive to packet delay -Request/Response pattern - Preserve payload content Interactive -Low error packet rate -Destination is not expecting the-data within a certain time -Preserve payload content, Background -Insensitive to packet delay, -Low error packet rate. Application example. Packet transmissio n priority. VoIP, Video telephony. 1st. Streaming video, Streaming audio Telnet, Chartroom, Web browser E-mail, SMS, File transfer. 2nd 3rd. 4th. From the QoS feature descriptions of UMTS traffic that mentioned above, the major distinguishing factor among these QoS features is the time sensitivity of packet transfer delay. Conversational traffic is the most delay sensitive, followed by streaming and interactive applications. Background traffic is the least delay sensitive. Moreover, compared to conversational and streaming traffic, interactive and background traffic 11.

(28) is sensitive to packet error rate since most of them are traditional Internet applications. In addition, interactive traffic bases on a request/response operation pattern, a long packet delay is not allowed. Thus, interactive traffic reaches a higher packet transmission priority than background traffic. The QoS features of UMTS traffic are summarized in Table 2 [16].. 2.3. UMTS with Differentiated Services (DiffServ) The UMTS has a number of service classes that require end-to-end QoS support. This requirement imposes on the design of the UMTS core network bearer service as a part of the UMTS hierarchical QoS architecture. This study presents queuing scheme and some scenario simulations for provisioning QoS in the UMTS core network based on the DiffServ model, a relatively simple but scalable IP-based QoS technology. This requires proper choices of QoS mapping, router configurations, and queuing scheme. Efficient queuing schemes are introduced for the UMTS core network gateway, particularly for the scheduling and buffer management schemes, to enhance QoS provisioning in UMTS. The effectiveness of this approach is illustrated by computer simulations. The next generation of mobile phones will be probably all-IP based enabling users to access Internet services. In order to make this possible a satisfactory QoS, at least equal to the fixed Internet, must be ensured. To achieve this goal an end-to-end QoS system must be constructed. Another fact is the dominance of IP over other technologies due, it is important to develop end-to-end IP QoS guarantees for the different applications over distinct access technologies [10]. This is particularly important for cellular wireless networks due to the ever growing expansion of mobile phone users. One way to contribute to this goal is to apply DiffServ QoS mechanisms to UMTS technology in order to model 12.

(29) an End-to-End QoS communication system. In particular, RED (Random Early Detection) queue management and PRI (Priority Queuing) or WRR (Weighted Round Robin Queueing) scheduling policies are enforced [20]. Different UMTS traffic classes are mapped into different DiffServ parameters [21, 22]. The performance of this architecture has been evaluated by simulation using NS2, assuming different network load scenarios.. 2.3.1. Integrated Services (IntServ) The integrated services (IntServ) scheme focuses on end-to-end individual packet flows. In this scheme is designed to provide a set of extensions to best-effort service model. The service level can be typically quantified as a minimum service rate, or a maximum tolerable end-to-end delay or loss rate. According to available resources, the network grants or rejects the flow requests. The admission control unit, the packet forwarding mechanisms and the Resource Reservation Protocol (RSVP) are the three major components of the IntServ architecture; these three components perform resources availability check, packet forwarding process in a router and bandwidth reservation jobs [23]. Since flow-based integrated services are supported in the IntServ scheme; routers require more complicated mechanisms to maintain control and packet forwarding states of all flows passing through them. For a router, it is a heavy load. From an viewpoint of implementation and operation, it is a significantly difficult job. Therefore, there exist scalability and manageability issues for the IntServ scheme [24]. The design principles of RSVP provides the end-to-end QoS guarantee for data stream and it is to reserve and maintain the resources at each node that is in the transmission path of flows. Figure 5 illustrates the best effort service and the IntServ. In this illustration, the term "traffic flow" is used in a loose sense and represents the source of traffic. In the best 13.

(30) effort service, all packets are lumped into a single mass regardless of the source of the traffic. In IntServ, individual flows are distinguished on an end-to-end basis.. Figure 5. Sketch map of Best effort and IntServ. 2.3.2. Differentiated Services (DiffServ) DiffServ attempts to accomplish the same goal as IntServ with better scalability, it is a simple model where traffic entering a network is first classified and possibly conditioned at the boundaries of the network, and then assigned to different behavior aggregates. Scalability is improved by moving per-flow states to edge routers and keeping information of only a few flow classes called Per-Hop-Behavior (PHB) in the core gateways. Packets are treated according to their PHB classes instead of individual flows. PHB can be performed by proper buffer management mechanisms such as RED in the gateways and fulfills scalability by performing all complicated QoS function, such as traffic classification, marking, metering, conditioning, shaping and per-flow traffic [22, 25]. The Diffserv control architecture definition shows in Figure 6, these functional elements are located in the ingress node of a DiffServ domain and in interior DiffServ-compliant nodes [26].. 14.

(31) Figure 6. DiffServ Control Architecture Classifier: Selects a packet in a traffic stream based on the content of some portion of the packet header. Meter: Checks compliance to traffic parameters (ie: token bucket) and passes results to the marker and shaper/dropper to trigger action for in/out-of-profile packets. Marker: Writes/rewrites the DSCP value Shaper: Delays some packets to be compliant with the profile.. The DiffServ has no dynamic admission control. Therefore, the network managers must make sure that enough resources are available for the agreed SLAs. DiffServ doesn’t support per-flow QoS guarantees to achieve scalability. It becomes challenging to still maintain QoS, especially for voice and video, which need per-flow guarantees [27]. The ways in which PHBs, edge functionality, and traffic profiles can be combined to provide an end-to-end service, such as a virtual leased line service [28].. 15.

(32) Chapter 3. Review of Queue Management Mechanisms The desired performance level and applying resource allocation policies in mobile communication network is a hard task. There is a need to express policies defined by the queuing scheme of each resource in a UMTS core network gateway. This need is becoming increasingly pressing in settings such as minimum limit, maximum limit, buffer size, type of packet, packet enqueueing probability…etc. The problem of queuing scheme in mobile communication network is becoming an important one.. 3.1 Passive Queue Management (PQM) In passive queue management (PQM), packets coming to a buffer are rejected only if there is no space in the buffer to store them and the senders have no earlier warning on the danger of growing congestion [29]. In this case all packets coming during saturation of the buffer are lost. The existing schemes may differ on the choice of packet to be deleted (end of the tail, head of the tail, random). During a saturation period all connections are affected and all reach in the same way, hence they become synchronized. The main drawbacks of PQM are summarized as follows [29]. Global synchronization When a drop-tail buffer is full, all of the incoming packets are dropped. Consequently, all the affected TCP connections try to recover those dropped packets at about the same time. This moment, all the connections simultaneously send large amount of packets to congest the buffer again. This phenomenon may seriously affect the link utilization and thus the overall network performance.. 16.

(33) Full-queue PQM is only activated after the buffer is full. The buffer occupancy may oscillate between empty and fullness. A traffic flow may experience large end-to-end variations. Lock-out Because of global synchronization phenomenon, some connections are always served first and the others are denied by PQM. The network resources are thus not distributed in a fair manner. This is called “lock-out”.. 3.2 Active Queue Management (AQM) To enhance the throughput and fairness of the link sharing, also to eliminate the synchronization, the Internet Engineering Task Force (IETF) recommends active algorithm of buffer management [30]. Active queue disciplines drop or mark packets before the queue is full. Typically, they operate by maintaining one or more drop/mark probabilities, and probabilistically dropping or marking packets even when the queue is short. Active Queue Management (AQM) can improve the performance of TCP, and has been recommended by the IETF for use in the routers of the mobile communication [29]. The goal of AQM is three folds. First, to improve throughput by reducing the number of packets dropped. This is achieved by keeping the average queue length small in order to absorb naturally occurring bursts without dropping packets. Second, AQM provides low delay to interactive services by maintaining a small average queue length. Third, AQM avoids the lock out phenomenon arising from tail drop [31, 32].. 17.

(34) 3.3. Random Early Detection (RED) Among various active queue management schemes, RED is probably the most popular studied. RED was proposed to improve the performance of TCP connections. As a queue management mechanism, it drops packets in the considered gateway buffer to adjust the network traffic behavior according to the queue size. Clearly, the configurable parameters of RED such as dropping probability and thresholds are critical to network performance, but the choice of these parameters remains more of an art than a science because of the complexity of the relationship between TCP/RED parameters and network performance. A few papers on mathematical models and parameter settings can be found in the literature [33, 34]. RED is an effective mechanism to control the congestion in the network routers. It also helps prevent the global synchronization in the TCP connections sharing a congested router and to reduce the bias against burst connections. It was an improvement over the previous proposals, such as Random Drop and Early Random Drop. Clark and Fang [35] have proposed the incorporation of DiffServ in the Internet by applying RED with different parameter setting to the “In” and “Out“ packets of the flows arriving at a router. To be able to apply RED mechanism in the DiffServ service, it is important to survey its queueing behavior. For example, to figure out any guarantees on throughput and delays one needs to survey these as a function of the RED parameters. It is useful to complement these efforts with an analytical study. Lin and Morris [36] have shown that the RED scheme doesn’t work particularly well when the queue is occupied by well-behaved TCP flows as well as greedy UDP flows at the same time. Misbehaving flows don’t back off even if their packets are dropped. The average value of the queue remains over Minth, causing drop from TCP flows that have already reduced their rate. 18.

(35) 3.4. Priority-based Queueing Priority-based queueing is a simple approach to provide DiffServ to different packet flow. Packets of different flows are assigned a priority level according to their QoS requirement. When packets arrive at the output link, they are first classified into different classes enqueued separately based on their priorities. Then, queues are served in order. The highest priority queue is served first before serving lower priority queues. Packets in the same priority class are serviced in a FIFO manner. But if a higher priority packet arrives while serving a lower priority packet, the server waits until complete the service of the current packet then goes back to serve the higher priority queue [37, 38]. Accordingly, higher priority queues will always be served before lower priority queues. If a high priority user offers more load than the link capacity of the output link, no packets can be transmitted from the lower priority queues. In the worst-case, all packets in lower priority queues will be discarded due to exceeding the transit delay bounds of the scheduler. The high-priority queue is always going to be given first priority, and that means traffic in the lower-priority queues can sit there for a long time. Figure 7 illustrates the priority-based queue, anytime a packet enters the high queue, the scheduler will stop transmitting any other queue's traffic and transmit the high-priority traffic. The packets in the other queues now have to wait their turn, if too many packets enter those higher-priority queues. The starvation or blocking problem always arises with high occurrence probability in the lower priority queue.. 19.

(36) Figure 7. Priority-based Queueing. The priority-based queuing discipline play a crucial role in the implementation of the DiffServ architecture where packets are classified into a number of traffic classes and handle with various priorities. Numberous research efforts have been made on performance analysis and evaluation of the priority queuing mechanism [29, 39], as well as its development and applications [40, 41].. 3.5. Weighted Round Robin (WRR) Queueing The traditional first-in-first-out (FIFO) droptail was initially the only queue management scheme in the network. It is simple and easy to implement in routers, however, it exacerbates the limitations of end-point congestion control schemes such as TCP. When a packet arrives and the queue is currently full, the incoming packet will be dropped. Droptail is the most widely used queue manage algorithm due to its simple implementation and relatively high efficiency. However, droptail has some weakness, such as the bad fairness sharing among TCP connection, and the throughput and link efficiency suffer severe degradation if congestion is getting worse [42]. Figure 8 and Figure 9 illustrate FIFO queueing and WRR diagrams.. 20.

(37) In WRR assigns a weight to each queue, and it then services each nonempty queue in proportion to its weight, in round-robin fashion. The packets in different queues are processed in turn. Thus the lowest-priority queue can be guaranteed of a minimum bandwidth. This avoids the case that the packets in the low priority queues cannot be served. WRR is optimal when using uniform packet sizes, a small number of flows, and long connections [43]. Lindemann and Thummler [44] have shown a QoS differentiation with WRR, but the main focus in the paper has been in balancing the resources between circuit switched voice calls and packet switched data calls.. Figure 8. First In First Out Queueing. Figure 9. Weighted Round Robin Queueing 21.

(38) Chapter 4. An Approach to Priority-based Queueing Scheme with Two Queueing Buffer Allocations It is impossible that a UMTS core network bandwidth can fully satisfy bandwidth requirements of all UMTS traffic. According to the QoS features of conversational, streaming, interactive, and background UMTS traffic and the packet transmission priority of each type of UMTS traffic, a priority-based queuing scheme with different queuing buffer allocations is proposed in this study to support a differentiated packet forwarding process for UMTS applications. The following subsections describe the processes of the enqueuing and dequeuing modules in details about the priority-based queuing scheme.. 4.1. Queueing Buffer Allocations in Priority-based Queueing Scheme Enqueuing and dequeuing modules are two major modules to handle packets enqueuing and dequeuing jobs in a queuing scheme. Usually, queuing buffer allocation in a queuing scheme affects packet forwarding processes in enqueuing and dequeuing modules directly [40, 45]. Considering packet forwarding in a DiffServ mechanism and packet forwarding starvation avoidance for four types of UMTS traffic, in [41] proposed assignment buffer access scheme and we use assignment buffer several queuing buffer allocation ideas, such as logical queuing buffer, guaranteed queuing buffer space, queuing buffer space dynamical allocation, and overflow queuing buffer space, to handle arrival packet enqueuing processes. With these queuing buffer allocation ideas, the proposed priority-based queuing scheme uses two queuing buffer allocations; one queuing buffer allocation is dynamic queuing buffer (DQB) allocation, the other is overflow queuing buffer (OQB) allocation. These two queuing buffer allocations will be described in details as the follows.. 22.

(39) 4.1.1. A Dynamic Queueing Buffer (DQB) Allocation In the DQB allocation, a queuing buffer is divided into four logical queuing buffers, conversational, streaming, interactive, and background. Each logical queue buffer is FIFO queue; it stores its corresponding UMTS packets. Each logical queue buffer is divided into two segments: guaranteed buffer and dynamic buffer. For one type of UMTS packets, a guaranteed buffer stores packets unconditionally if space is available. Since each type of UMTS traffic has its own guaranteed buffer to enqueue arrival packets; it might reduce a possibility of a UMTS packet enqueuing starvation, especially for UMTS applications with lower packet transmission priorities. Moreover, a guaranteed buffer size depends on a packet transmission priority of its corresponding UMTS traffic. Usually, a logical queuing with a high packet transmission priority, more guaranteed buffer space would be allocated; this would be helpful for UMTS packets with higher packet transmission priorities to be enqueued easily. When no space is available in a guaranteed buffer, a shared buffer will be allocated conditionally to store an arrival packet [46, 47]. A RED-based packet enqueuing process is invoked by the proposed enqueuing module. According to available space in queuing buffer and related parameters settings, the DQB allocation process is invoked to decide to enqueue or drop arrival packets. As an arrival packet is allowed to store in queuing buffer; it will be enqueued into its corresponding dynamic buffer [48, 49]. A dynamic buffer is the buffer space will be appended to a logical queuing buffer dynamically when one arrival packet is enqueued into it. A size setting of a dynamic buffer of each logical queuing buffer depends on its corresponding packet transmission priority. A logical queuing buffer with a high packet transmission priority might have a larger dynamic buffer than a logical queuing buffer with a low packet transmission priority. Figure 10 shows the diagram of the queuing buffer allocation in the DQB allocation. 23.

(40) Figure 10. A diagram of queuing buffer with the DQB allocation. 4.1.2. An Overflow Queueing Buffer (OQB) Allocation Like the DQB allocation, a FIFO logical queuing buffer idea also is adopted in the OQB allocation. In the OQB allocation, five FIFO logical queuing buffers are used to enqueue arrival UMTS packets; four logical queuing buffers are guaranteed buffers and one logical queuing buffer is a shared buffer. The four guaranteed logical queuing buffers are corresponding to four types of UMTS traffic separately; an arrival packet will be enqueued into its corresponding guaranteed logical queuing buffer unconditionally when space is available to enqueue one packet. Allocations of these four guaranteed logical queuing buffers let four types of UMTS traffic can receive their queuing buffer space to enqueue their arrival packets; it can avoid packet enqueuing starvation. Moreover, for enhancing an enqueuing possibility of UMTS packet with a higher packet transmission priority, these four logical queuing buffers depend on their packet transmission priorities to allocate their queuing buffer space. An overflow logical queuing buffer is one shared buffer; it is shared by all UMTS packets. A size of overflow logical queuing buffer is a difference between a physical queuing buffer size and 24.

(41) four guaranteed logical queuing buffers. For a UMTS arrival packet, if no space is available to accommodate one packet in its corresponding logical queuing buffer; the overflow queuing buffer will base on a RED-based packet enqueuing process to determine whether an arrival packet can be enqueued into the overflow queuing buffer or not. Usually, UMTS traffic with a higher packet transmission priority receives better settings in its corresponding RED-parameters to allow more packets with the same packet transmission priority to be enqueued into the overflow logical queuing buffer more easily. Figure 11 shows a diagram of the OQB allocation.. Figure 11. A diagram of queuing buffer with the OQB allocation. 4.2. A Priority-Based Enqueueing Module A priority-based packet enqueuing module [41, 50, 51] is adopted to handle arrival UMTS packets enqueuing process among several logical queuing buffers in the proposed queuing scheme. Since two types of queuing buffers are allocated to enqueue arrival packets; a two-stage packet enqueuing process is adopted to process an arrival packet enqueuing job in the priority-based enqueuing module. In the first stage packet enqueuing process, a FIFO method is used for a guaranteed buffer; 25.

(42) an arrival packet can be enqueued into a guaranteed buffer only if space is available in its corresponding guaranteed buffer. Otherwise, a RED-based packet enqueuing process is invoked in the second stage process to decide to enqueue or drop an arrival packet. Figure 12 shows the pseudo code of the two-stage packet enqueuing process. A two stage priority-based enqueueing process It bases on packet transmission priorities of four types of UMTS traffic to set the parameters: guaranteed buffer sizes, minimum limits, maximum limits and packet enqueueing probabilities, which are used in the priority-based enqueueing module If. (queueing queue buffer space is available) { /* to invoke the first stage */ if (enqueued packets size < guaranteed buffer size) { To enqueue an arrival packet into a guaranteed buffer in its corresponding logical queue } else { /* to invoke the second stage */ To invoke the RED-based packet enqueueing process } } else { /* the queueing buffer is full */ To drop an arrival packet } Figure 12. A pseudo code of a two stage priority-based enqueuing process. The priority-based packet enqueuing module based on packet transmission priority of each type of UMTS traffic to set the parameters which are used in the two-stage packet enqueuing process. These parameters include guaranteed buffer size of each logical queue buffer and four groups of RED-alike parameters [33, 34]. Each group of RED-alike parameters consists of minimum limit, maximum limit, and packet enqueuing probability. Generally speaking, each logical queuing buffer depends on its packet transmission priority to receive corresponding guaranteed buffer size and RED-alike parameter settings. A logical queuing buffer with a higher packet transmission priority always receives more favorable parameter settings than logical queuing buffer with a lower packet transmission priority. It is easier for an arrival UMTS packet with a higher packet transmission priority to be enqueue 26.

(43) into its corresponding logical queuing buffer. By way of corresponding parameter settings, a DiffServ behavior can be supported by the proposed enqueuing module. Since two queuing buffer allocations are adopted in the priority-based enqueuing module to handle an arrival packet enqueuing process, there exists some differences in the proposed packet enqueuing process. The proposed packet enqueuing process with the two queuing buffer allocations will be described in details as the follows.. 4.2.1. A Priority-based Packet Enqueueing Module with a DQB Allocation With the DQB allocation, two segments are allocated in each logical queuing buffer; one is a guaranteed buffer and the other is a dynamic buffer, it is a shared buffer. For enqueuing arrival packets into guaranteed buffer and dynamic buffer in a logical queuing buffer, the two-stage packet enqueuing process uses two different measures to process an arrival packet enqueuing job. In the first stage packet enqueuing process, a FIFO method is used for a guaranteed buffer; an arrival packet can be enqueued into a guaranteed buffer only if space is available in its corresponding guaranteed buffer. Otherwise, a RED-based packet enqueuing process is invoked in the second stage process to decide to enqueue or drop an arrival packet. Packet enqueuing process in the second stage handles a dynamic buffer space allocation for an arrival packet. Since a dynamic buffer space allocation depends on available space of a physical queuing buffer; it is easier for an arrival packet to receive dynamic buffer space when more space is available in a physical queuing buffer. Thus, the RED-based packet enqueuing process bases on one group of RED-alike parameters which are corresponding to an arrival packet and available space in physical queuing buffer to determine whether dynamic buffer 27.

(44) space can be allocated or not. If dynamic buffer space can be allocated for an arrival packet; an arrival packet would be enqueued into its corresponding dynamic buffer. Allocated dynamic buffer space will be appended to the end of one logical queuing buffer which is corresponding to the arrival packet. Otherwise, an arrival packet will be dropped by the RED-based packet enqueuing process immediately.. Figure 13. A diagram of arrival packets enqueueing process with the DQB allocation. In the RED-based packet enqueuing process, two thresholds [31, 33], minimum limit and maximum limit, about a physical queuing buffer utilization are use to decides whether an arrival packet can receive dynamic buffer space allocation unconditionally or to be dropped immediately. Dynamic buffer space can be allocated for an arrival packet unconditionally when length of enqueued packet is less than minimum limit. An arrival packet will be dropped immediately when length of enqueued packet is more than maximum limit. Moreover, Dynamic buffer space can be allocated for an arrival packet conditionally when length of enqueued packet is more than minimum limit and it is less than maximum limit. The diagram of arrival packets enqueueing process with the DQB allocation is shown in Figure 13 and Figure 14 shows the 28.

(45) pseudo code of the RED-based packet enqueuing process with the DQB allocation. A RED-based packet enqueuing process with the DQB allocation if (physical queuing buffer space is available) { if (the corresponding logical buffer size of an arrival packet ＜ its minimum limit) { To allocate dynamic buffer space with the DQB allocation and enqueue an arrival packet into the allocated dynamic buffer space To append to the allocated dynamic buffer space to one logical queuing buffer which is corresponding to the arrival packet } else { if (the corresponding logical buffer size of an arrival packet ≧ its maximum limit ) { To drop an arrival packet immediately } else { To generate a random probability based on a uniform distribution if (the random probability ≦ the packet enqueuing probability ) { To allocate dynamic buffer space with the DQB allocation and enqueue an arrival packet into the allocated dynamic buffer space To append to the allocated dynamic buffer space to one logical queuing buffer which is corresponding to the arrival packet } else { To drop an arrival packet } } } else { /*no space available in queuing buffer */ To drop an arrival packet immediately } Figure 14. A pseudo code of the RED-based packet enqueuing process with the DQB allocation. 29.

(46) 4.2.2. A Priority-based Packet Enqueueing Module with an OQB Allocation Like the priority-based enqueuing module with the DQB allocation, a two-stage packet enqueuing process is adopted in this enqueuing module, too. The process procedure of two-stage packet enqueuing process in this enqueuing module is same as the process procedure of two-stage packet enqueuing process with a DQB allocation. However, there exist differences in queuing buffer allocation between the DQB allocation and the OQB allocation; this causes that the two-stage packet enqueuing process has different process target with the different queuing buffer allocation. Four logical queuing buffers which are corresponding to four types of UMTS traffic are guaranteed buffers; they are the guaranteed buffer space to store arrival UMTS packets with the OQB allocation. Guaranteed buffer space allocation of each type of UMTS traffic depends on its packet transmission priority. A UMTS application with higher packet transmission priority receives more guaranteed buffer space allocation. Differentiated guaranteed buffer space allocations exist among four types of UMTS traffic. FIFO method is used in the first stage process to handle arrival UMTS packets enqueuing job in guaranteed buffers. Arrival packets can be enqueued only if space is available in their corresponding guaranteed buffer; otherwise, a RED-based packet enqueuing process will be invoked in the second stage process to handle arrival packet enqueuing job in an overflow logical queuing buffer. The second stage packet enqueuing process is a RED-based packet enqueuing process; it bases on the overflow queuing buffer’s capacity and type of an arrival UMTS packet to handle arrival packet enqueuing job in the overflow logical queuing buffer. Like the RED-based packet enqueuing process with the DQB allocation, each type of UMTS traffic 30.

(47) has its corresponding RED-alike parameters minimum limit, maximum limit, and packet enqueuing probability, to calculate an enqueued probability of an arrival packet in this stage process. And, the settings of four groups of RED-alike parameters are corresponding to their packet transmission priorities. A packet with a higher packet transmission priority receives better settings in its corresponding RED-alike parameters and it can be enqueued into the overflow logical queuing buffer more easily. With proper parameters settings, a differentiated packet enqueuing behavior can be supported in the proposed enqueuing module with the OQB allocation. The diagram arrival packets enqueueing process with the OQB allocation is shown in Figure 15.. Figure 15. A diagram of arrival packets enqueueing process with the OQB allocation. Since there exists differences in a shared buffer allocation between the OQB allocation and the DQB allocation, the RED-based packet enqueuing process with the OQB allocation is a little different from to the RED-based packet enqueuing process with the DQB allocation. Essentially, these two RED-based packet enqueuing processes base on three parameters, minimum limit, maximum limit and packet enqueuing probability, to decide whether an arrival packet can be enqueued into a shared buffer or not. In the OQB allocation, an overflow logical queuing 31.

(48) buffer is the only shared buffer; it stores all enqueued packets which are allowed by the RED-based packet enqueuing process with the OQB allocation. A FIFO method is used to process packet enqueuing and dequeuing jobs in an overflow logical queuing buffer. A packet enqueuing sequence is determined by packet enqueuing time and four types of enqueued packets are mixed in an overflow logical queuing buffer. Figure 16 shows the pseudo code of the RED-based packet enqueuing process with the OQB allocation. A RED-based packet enqueueing process with the OQB allocation if (physical queueing buffer space is available) { if (enqueued packet size of the overflow logical buffer size ＜ a minimum limit of an arrival packet) { To enqueue an arrival packet into the overflow logical queueing buffer } else { if (enqueued packet size of the overflow logical buffer size ≧ its maximum limit of an arrival packet) { To drop an arrival packet immediately } else { To generate a random probability based on a uniform distribution if (the random probability ≦ the packet enqueueing probability ) { To enqueue an arrival packet into the overflow logical queueing buffer } else { To drop an arrival packet } } } else { /*no space available in queueing buffer */ To drop an arrival packet immediately } Figure 16. The pseudo code of a RED-based packet enqueueing process with the OQB allocation. 32.

(49) 4.3. A WRR Packet Dequeuing Module A packet dequeuing module in a queuing scheme performs packets forwarding from its queuing buffer to packets’ next hop gateways. Since all enqueued packets are stored in several logical queuing buffers no matter with the DQB allocation or the OQB allocation. Therefore, for supporting a DiffServ in packet deuqueuing process and avoiding packet dequeuing starvation, especially for packets with a lower transmission priority, a WRR idea is applied to propose a sequential-WRR (SWRR) dequeuing scheme in this packet dequeuing module [43]. A diagram of the SWRR dequeuing scheme with the DQB allocation is shown Figure 17 and a diagram of the SWRR dequeuing scheme with the OQB allocation is shown Figure 18.. Figure 17. A diagram of the SWRR dequeuing module with the DQB allocation. In the SWRR scheme, each logical queuing buffer depends on its corresponding packet transmission priority to receive its deserved packet dequeuing weight in a weighted packet dequeuing round robin cycle. In a weighted packet dequeuing round robin cycle, each logical queuing depends on its deserved dequeuing weight to receive packet dequeuing turns in an assigned dequeuing sequence. One packet will be dequeued 33.

(50) from one logical queuing buffer when the logical queueing buffer receives a packet dequeuing turn. A dequeuing weight allotment bases on packet transmission priorities of logical queuing buffers and a number of logical queuing buffers to allot a corresponding packet dequeuing turns to each logical queuing buffer. Since the DQB allocation and the OQB allocation have different logical queuing buffers; a logical queuing buffer with the same packet transmission priority will receive different packet dequeuing turns in these two queuing buffer allocations. In the SWRR scheme, at least one packet, possibly more, will be dequeued from each logical queuing buffer in a packet dequeuing cycle. Usually, packets with a high transmission priority are more easily dequeued than packets with a low transmission priority. The packet dequeuing turns received by logical queuing buffers in the DQB allocation and the OQB allocation are shown in Table 3.. Figure 18. A diagram of the SWRR dequeuing module with the OQB allocation. 34.

(51) Table 3. Packet dequeuing turns received by all logical queuing buffers with the DQB / OQB allocations in a packet dequeuing cycle. Packet dequeuing turn number the DQB allocation the OQB allocation 4 5 3 4 2 2 1 1 0 3 10 15. Logical queuing buffer Conversational Streaming Interactive Background Overflow Dequeueing cycle. The SWRR depends on a congruence equation to handle a dequeuing turn switching a process among all logical queuing buffers. Two variables, a packet dequeuing counter and a packet dequeuinging cycle, are the core parameters in the congruence equation. The packet dequeuing counter increments when one packet dequeued from one logical queuing buffer. The packet dequeuinging cycle is the sum of the assigned dequeuing turns received by all logical queuing buffers in a weighted packet dequeuing round robin cycle. Then, with a mapping relationship between a congruence which is calculated by the congruent equation and a packet dequeuing sequence among logical queuing buffers, dequeuing turns will be switched among all logical queuing buffer dispersedly and differential dequeuing turns can be received by logical queuing buffers. a congruence = packet dequeueing counter mod packet dequeueing cycle. … (2). packet dequeueing cycle = total logical queueing buffer number. ∑. packet dequeueing turns i. ….(3). i =1. In additions, as one logical queueing buffer receives a packet dequeueing turn and there is no packet in the logical queueing buffer to be dequeued; a packet dequeueing turn transfer (PDTT) procedure will be invoked by the SWRR to accelerate packets dequeueing process. The 35.

(52) PDTT procedure will try to find one logical queueing buffer which has a higher packet transmission priority and has one packet can be dequeued. If one logical queueing buffer is found; it will receive a packet dequeueing turn which is transferred from another logical queueing buffer. With the PDTT procedure, one logical queueing buffer with a higher packet transmission priority always can receive a transferred packet dequeueing turn to dequeue one packet from itself. Thus, Figure 19 shows the flowchart of PDTT procedure which will be enhanced a differentiated packet dequeueing behavior. A flowchart of the proposed weighted round robin packet dequeueing module is shown in Figure 20.. Figure 19. Flowchart of dequeueing turn transfer PDTT procedure. 36.

(53) Figure 20. Flowchart of the WRR packet dequeueing module. 37.