Below we provide the pseudo-code for our proposed SNG routing algorithm. In a nutshell, whenever there is a downloading request (DLRequest == true) issued, the SNG routing daemon computes an optimized route based on discovered neighborhood
infor-mation, which includes network configurations, available gateways, attainable downlink capacities from the IP network, and estimated ad hoc link rates. The ultimate goal is to maximize the perceived downlink data rate for a requesting client that participates in the synergized framework (SF).
Algorithm 1 SNG Routing Algorithm
1: while (!exit) do
2: Periodically estimate neighboring link capacities;
3: Periodically exchange SNG C-Table;
4: if DLRequest == true then
5: Construct the network graph G = (V , E);
6: for each gateway candidate vg ∈ Vg do 7: Solve LP (input: ckij, Iijk, K, Vg, Vm, E);
8: output: fijk, λ;
9: end for
10: Send GREQ to selected gateway node with the max λ;
11: Send RREQ to all relaying nodes;
12: Refresh all link & gateway capacities;
13: Perform downloading;
14: end if 15: end while
Chapter 4
Performance Evaluation
To validate the performance of the proposed routing mechanism in the hybrid net-work, we implement our SNG protocol in the ns-2 simulator. The IEEE 802.11 MAC protocol with RTS/CTS four-way handshaking is used, and 3 802.11b non-overlapping channels are simulated. Default transmit power and Two-Ray Ground propagation model are adopted, leading to 250m transmission distance and 550m interference range. For each client running the SNG routing optimization, the lpsolve tool is utilized for the compu-tation [1]. As presented in Section 3.1, SNG does not intend to collect a global network information. Instead, a TTL field is used to limit how far the C-Table can propagate.
In the simulations, we use the parameter h to indicate how many hops the C-Table can travel. Two other routing mechanisms, Greedy and On-demand (introduced in [16]), are also implemented for comparison with SNG. The Greedy protocol searches for proxy client with better downlink quality, starting from the immediate neighbors, in a greedy man-ner until no better neighbor can be found. On the other hand, the On-demand protocol selects the proxy with the highest downlink capacity within h-hop neighborhood.
A Level V (300 Kbps) Level IV (800 Kbps) Level III (1 Mbps) Level II (2 Mbps) Level I (5 Mbps)
Figure 4.1: Simulation environment with 45 nodes randomly deployed in a 1500m × 1500m network topology, along with corresponding Internet connection status, interface (channel) configurations, and ad hoc wireless link rates.
4.1 Impact of Network Information Scope
In this section, we investigate the impact of parameter h, and compare different routing strategies with respect to obtained downlink throughput. Fig. 4.1 shows the simulated network environment. Colored nodes represent potential gateways, which are classified into five levels in terms of downlink capacities. Channel (interface) configurations and estimated link rates are also illustrated in Fig. 4.1. In order not to further complicate the network environment, symmetric links (with equal link rate in both directions) are modeled in the simulations (note that this simplification does not affect the performance justification). We generate a mobile client A, which has two 802.11 interfaces operating on Channel 1 and 2 respectively and one cellular interface connecting to the IP network, to observe the attainable downlink throughputs at different time snapshots (t0-t6). As client A roams across the network, its cellular channel qualities vary depending on its locations.We implement four approaches for client A to obtain downlink services: Without Relay (using its own cellular connections), Greedy, On-demand, and our SNG.
t0 t1 t2 t3 t4 t5 t6
Client A Throughput (Kbps)
SNG (h=3)
Figure 4.2: Throughput performance obtained by using different routing strategies as client A roams across the simulated network.
Fig. 4.2 shows that SNG outperforms other strategies when reasonable neighborhood information (h ≥ 2) can be utilized for route optimization, while Without Relay provides the lowest downlink throughput due to no cooperation with other clients. We also exper-iment on SNG (h = 4, h = 5), but the throughput improvement is insignificant compared to SNG (h = 3), thus omitted from the figure. The results indicate that SNG only needs a moderate amount of neighborhood knowledge to outperform other strategies. In or-der to have a better unor-derstanding of how each routing mechanism determines the best route, we compile the downlink paths selected by respective protocol at t2, t3, and t6 in Fig. 4.3. Interestingly, the best route (leading to the highest downlink throughput) does not always involve the best gateway (SNG at t2, t3), or the shortest hop distance (SNG at t6). Moreover, at t2, Greedy gives lower throughput with shorter route (2-hop) than the On-demand strategy with longer route (3-hop), due to weak link quality over Ch2 (0.4 Mbps) used by Greedy. This phenomenon also reveals that the hop distance factor alone cannot act as the single metric for a good route. In addition, downlink throughput can
A
Figure 4.3: Downlink flow paths selected by different routing strategies at respective time snapshot. Note that selecting Internet proxy (gateway) with the highest downlink rate does not necessarily yield the best throughput for client A, since the communication bottleneck may exist in the ad hoc network domain.
be further increased by enabling multi-path packet delivery (SNG at t6). From the above observations, we conclude that a good routing protocol in such a hybrid network should take various factors into a unified consideration, thus validating the SNG optimization strategy.
4.2 Importance of System Load Balancing
Another essential problem for the heterogeneous multi-hop network is the system load distribution. We investigate this issue by observing the aggregate throughput produced by respective routing technique. Fig. 4.4 shows the simulated network, with five potential gateways, and up to 15 user requests made in order. The aggregate network throughputs under different routing strategies are plotted in Fig. 4.5. The SNG protocol is able to yield significantly higher throughput than the other two routing mechanisms. Since SNG refreshes gateway capacities and ad hoc link rates to reflect current bandwidth
U11 Ui User Requests
Other Nodes
Figure 4.4: Network configuration with 30 clients and 5 potential proxies (gateways) randomly placed in a 1500m × 1500m topology.
occupied by existing users, the routing process has better knowledge to distribute the downloading requests correctly. In addition, we analyze the gateway utilization status, shown in Fig. 4.6, under different routing strategies. With the same user requesting patterns (U1 request, followed by U2 request, followed by U3 request, and so on), we observe that G1 is under-utilized in Greedy and On-demand. In fact, G1 is selected by several users, such as U3and U4, as downlink gateway. However, because of ad hoc channel contentions, U3 and U4 seldom get the chance to use G1 capacity. On the other hand, SNG avoids such adversary effect by distributing traffic to other non-interfering channels, leading to high G1 utilization. Two important design issues are revealed from the above experiments. First, both gateway and ad hoc link capacities should be refreshed to reflect up-to-date bandwidth allocation. Second, channel diversity should be leveraged.
0 2 4 6 8 10 0
0.5 0.1 1.5 2.0 2.5 3.0 3.5 4.0 4.5
Number of User Requests
Aggregate Network Throughput (Mbps)
SNG (h=3) On−demand (h=2) Greedy
Figure 4.5: Aggregate network throughput increases as user demands grow, and satu-rates at varying points under different routing strategies (without limiting per-user traffic amount).
4.3 Restricting Per-user Traffic to Accommodate More Users
From Fig. 4.5, we further analyze the distributions of per-user traffic flows under different routing strategies. Fig. 4.7 indicates the unevenly distributed flow occupancy, leading to unfair network capacity sharing among users. Only 6, 8, and 5 users are effectively served (with non-zero percentage) using the SNG, On-demand, and Greedy mechanisms, respectively. In this set of experiments, we impose a 500 Kbps limitation on per-user downlink flow allowed by available gateway capacity. The results are shown in Fig. 4.8 with slightly reduced throughputs (compared to Fig. 4.5). By imposing the per-user flow restriction, the aggregate throughputs gradually reach the saturation points (in contrast to the sharp saturations in Fig. 4.5). Fig. 4.9 illustrates the improved flow distributions, where more users can be supported by the available system capacity. Due to
0
Number of User Requests
saturated saturated
saturated saturated
saturated saturated
Figure 4.6: Proxy (gateway) capacity utilization status under Greedy, On-demand, and our SNG routing strategies, respectively.
better balanced bandwidth allocations, 12, 9, and 9 effective requests can now be served by the SNG, On-demand, and Greedy protocols, respectively.
Based on a series of simulative experiments targeted on the heterogeneous multi-hop networking environment, among the three routing techniques, our proposed SNG protocol is demonstrated to be capable of offering the best downlink rate through its optimized routing intelligence.
SNG (h=3) Saturated at 4.40 Mbps Saturated at 1.86 Mbps Saturated at 2.06 Mbps
U1
Figure 4.7: Unevenly distributed per-user traffic flow occupancy among saturated throughput obtained by respective routing strategy.
0 5 10 15
Number of User Requests
Aggregate Network Throughput (Mbps)
SNG (h=3) On−demand (h=2) Greedy
Figure 4.8: Aggregate network throughput against number of user requests (with 500 Kbps restriction on per-user traffic amount). More user requests can be accommodated by the overall system capacity.
U1 Saturated at 4.12 Mbps Saturated at 1.70 Mbps Saturated at 1.92 Mbps
U1
Figure 4.9: Improved per-user traffic flow distributions to allow more user requests at the cost of slightly reduced system saturated throughputs (compared with Fig. 4.7).
Chapter 5
Conclusion and Future Work
In this thesis, we propose a synergized framework (SF), and design a network-layer SNG routing protocol to enable the proposed communication model. The SNG protocol determines the best route for a downloading request in an optimized manner. Simu-lations show that the SNG mechanism is able to outperform other routing strategies based on moderate amount of neighborhood information without incurring much com-putation/communication overhead. The results encourage us to initiate a prototyping project. In the near future, we plan to set up a heterogeneous multi-hop networking testbed, containing various wireless devices, and implement the SNG protocol to perform packet routing. With the availability of open-source Android-based platforms, we may be able to incorporate such routing intelligence into smart phones as well, making them even smarter. Those prototyping experiences will be reported in our future publication.
Although we only consider the Internet downloading traffic in the current SNG routing process, the optimization strategy can be easily extended to support peer-to-peer com-munication within the ad hoc network domain, finding the best route between arbitrary source and destination node pair. Such extension will also be directed into our future study.
Bibliography
[1] lp-solve: a Mixed Integer Programming (MIP) solver. http://sourceforge.net/
projects/lpsolve/.
[2] Must-Have Reference for Wireless Communication - Understanding Wireless Tele-com and Data Communications Terminology. http://www.eu.anritsu.com/
musthaveguides, December 2008.
[3] G. N. Aggelou and R. Tafazolli. On the Relaying Capability of Next-Generation GSM Cellular Networks. IEEE Personal Communications, 8(1):40–47, February 2001.
[4] F. O. Akg¨ul and M. O. Sunay. Enhancing 3G High Speed Wireless Data Performance Through Utilization of 3G—WLAN System Cooperation: A Network, Cross-Layer Approach. Wireless Personal Communications, 43(1):89–106, October 2007.
[5] D. Cavalcanti, D. Agrawal, C. Cordeiro, B. Xie, and A. Kumar. Issues in Integrating Cellular Networks, WLANs, and MANETs: A Futuristic Heterogeneous Wireless Network. IEEE Wireless Communications, 12(3):30–41, June 2005.
[6] L.-J. Chen, T. Sun, G. Yang, M. Y. Sanadidi, and M. Gerla. Ad Hoc Probe: Path Ca-pacity Probing in Wireless Ad Hoc Networks. In Proc. IEEE Int’l Conf. on Wireless Internet (WICON), pages 156–163, July 2005.
[7] Z. Dziong, F. Khan, K. Medepalli, and S. Nanda. Wireless Internet Access Using IS-2000 Third Generation System: A Performance and Capacity Study. Wireless Networks, 8(4):325–336, July 2002.
[8] H.-Y. Hsieh and R. Sivakumar. Performance Comparison of Cellular and Multi-Hop Wireless Networks: A Quantitative Study. In Proc. ACM SIGMETRICS, pages 113–122, June 2001.
[9] H.-Y. Hsieh and R. Sivakumar. On Using the Ad-Hoc Network Model in Cellular Packet Data Networks. In Proc. ACM MobiHoc, June 2002.
[10] I. Ioannidis, B. Carbunar, and C. Nita-Rotaru. High Throughput Routing in Hybrid Cellular and Ad Hoc Networks. In Proc. IEEE Int’l Symposium on a World of Wireless Mobile and Multimedia Networks (WoWMoM), 2005.
[11] K. Jang, M. Han, S. Cho, H.-K. Ryu, J. Lee, Y. Lee, and S. B. Moon. 3G and 3.5G Wireless Network Performance Measured from Moving Cars and High-Speed Trains. In Proc. ACM Int’l Workshop on Mobile Internet Through Cellular Networks (MICNET), pages 19–24, September 2009.
[12] X. Li and L. Shen. Performance Analysis in C3G-A System. In Proc. ACM Int’l Conf. on Wireless Communications and Mobile Computing (IWCMC), June 2009.
[13] X. J. Li and P. H. J. Chong. Performance Analysis of Multihop Cellular Network with Fixed Channel Assignment. Wireless Networks, February 2006.
[14] T.-Y. Lin, W.-H. Tam, K.-L. Fan, and Y.-C. Tseng. Resource Planning and Packet Forwarding in Multi-Radio, Multi-Mode, Multi-Channel, Multi-Rate (M4) Wireless Mesh Networks. Elsevier Computer Communications, 31(7):1329–1342, May 2008.
[15] Y.-D. Lin and Y.-C. Hsu. Multihop Cellular: A New Architecture for Wireless Communications. In Proc. IEEE INFOCOM, pages 1273–1282, November 2000.
[16] H. Luo, R. Ramjee, P. Sinha, L. E. Li, and S. Lu. UCAN: A Unified Cellular and Ad-Hoc Network Architecture. In Proc. ACM Int’l Conf. on Mobile Computing and Networking (MobiCom), pages 353–367, September 2003.
[17] P. Maill´e. Allowing Multi-Hops in Cellular Networks: an Economic Analysis. In Proc. ACM MSWiM, October 2005.
[18] B. S. Manoj, K. J. Kumar, C. F. D, and C. S. R. Murthy. On the Use of Multiple Hops in Next Generation Wireless Systems. Wireless Networks, 12:199–221, 2006.
[19] Motohiro Tanno. Standardizations for LTE/LTE-Advanced and Future Broadband.
http://apwcs2009.yu.ac.kr/, August 2009.
[20] T. Rouse, S. McLaughlin, and H. Haas. Coverage-Capacity Analysis of Opportunity Driven Multiple Access (ODMA) in UTRA TDD. In Proc. IEEE Int’l Conf. on 3G Mobile Communication Technologies, pages 252–256, March 2001.
[21] S. Tanaka, H. Ishii, T. Sao, U. Iizuka, and T. Nakamori. HSDPA Throughput Per-formances Using an Experimental HSDPA Transmission System. NTT DoCoMo Technical Journal, 6(4):19–28, March 2005.
[22] H.-Y. Wei and R. D. Gitlin. Two-Hop-Relay Architecture for Next-Generation WWAN/WLAN Integration. IEEE Wireless Communications, 11(2):24–30, April 2004.
[23] H. Wu, C. Qiao, S. De, and O. Tonguz. Integrated Cellular and Ad Hoc Relaying Systems: iCAR. IEEE Journal on Selected Areas in Communications, 19:2105–2115, October 2001.
[24] X. Wu, S.-H. G. Chan, and B. Mukherjee. MADF: A Novel Approach to Add an Ad-Hoc Overlay on a Fixed Cellular Infrastructure. In Proc. IEEE WCNC, 2000.
[25] X. Wu, S.-H. G. Chan, B. Mukherjee, and B. Bhargava. MADF: Mobile-Assisted Data Forwarding for Wireless Data Networks. Journal of Communications and Net-works, 6(3), September 2004.
[26] Y. Wu, K. Yang, and J. Zhang. An Adaptive Routing Protocol for an Integrated Cellular and Ad-Hoc Network with Flexible Access. In Proc. ACM Int’l Conf. on Wireless Communications and Mobile Computing (IWCMC), pages 263–268, 2006.
[27] A. N. Zadeh, B. Jabbari, R. Pickholtz, and B. Vojcic. Self-Organizing Packet Radio Ad Hoc Networks with Overlay (SOPRANO). IEEE Communications Magazine, June 2002.
自 傳
我來自嘉義的一個小康家庭，家境不算富裕。家中有五名成員，父母親、二 個妹妹、我。小時候在父母親的諄諄教誨之下，完成了小學、國中及高中的學業。
高中畢業後大學沒考上、重新補習一年考上義守大學資訊工程系，並由那時開始 接觸了電腦，並慢慢對電腦產生興趣，因此在學校期間，參加了電腦研習社，多 方的學習電腦技能。
大學畢業後即入伍當兵，退伍後第一份工作是到神達電腦公司當 BIOS 軟體 助理工程師，但因發現自己不是很有興趣在 BIOS 開發上，所以三個月後就換了 另一份工作，總算找到自己有興趣的工作，在廣聯國際股份有限公司當測試工程 師，後來一直對測試很有興趣，所這份工作做的很不錯，之後在同事介紹下到光 寶公司的通訊事業群當手機的測試工程師，負責 Nokia CDMA 案子的 field trial，去過印度、美國、印尼等國家。後來因光寶結束通訊事業群，所以跟著 在光寶時的主管一起轉職到華冠通訊股份有限公司，一樣負責產品的測試驗證工 作。
在華冠這期間公司陸續參與幾個 project， 如一開始的 3G feature phone，
及後來的 Windows mobile 手機，負責了 Windows mobile 6.1 LTK 的測試認證 工作，並且第一支通過 Windows Mobile 6.1 LTK 認證的手機，此手機是跟 Asia O2 合作的案子。另外也有跟倚天合作的有 GPS 的 Windows Mobile 的手機，並且 去過香港 field trial。並於今年的五月內轉到 protocol 部門，參與 Qualcomm 的案子，負責 Qualcomm 平台的 protocol 測試，所以在手機通訊這塊領域已累積 相當的經驗。
對於職場工作上，自己所負責的工作、任務都會盡心去完成，並且從中一直 學習。並且為了希望能多學習加強自己技能，在華冠期間利用時間再進修，考上 交通大學電信在職研究所，並且已順利畢業。期望將來能在職場上更能發揮所長。