Transcript Document
A SPEED ADAPTIVE MOBILE INTERNET PROTOCOL OVER WIRELESS LOCAL A REA NETWORK Final Defense Presentation Jun Tian Nov. 15th 2005 Contents 1. Instruction 2. Related Works 3. Performance of MIP over WLAN at Different Speeds 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latencies 5. A Speed Adaptive MIP and Its Performance Evaluation 6. Summary and Future Works 1 Motivation Trend 1- more and more mobile users The worldwide number of Internet users will reach nearly 1 billion in mid 20 05. The U.S. continues to lead with over 185M Internet users forecasted for ye ar-end 2004. There is little Internet user growth in the developed countries, b ut in the next five years many Internet users will be supplementing PC Intern et usage with Smartphone and mobile device Internet usage. Internet usage is growing strongly in China, which surpassed Japan for second place in 2003. T he growth of Internet users will continue in the developing countries for anot her decade. “Much of future Internet users growth is coming from populous countries suc h as China, India, Brazil, Russia and Indonesia. These countries will also see s trong growth of wireless web usage and for many new Internet users the cell phone will be their only Internet access device. ” InternetIndustry Users byAlmanac Regions -- September 3, 2004 - Computer Source: eTForecasts 1 Motivation Trend 1- more and more mobile users Wireless Internet Usage and Projections Year-End 2001 2004 2007 533 945 1,460 16.0% 41.5% 56.8% Internet Users (millions) 149 193 236 Wireless Internet User Share 4.5% 27.9% 46.3% 115 357 612 34.8% 50.9% 60.4% 126 208 290 13.9% 49.6% 67.0% Worldwide Internet Users (millions) Wireless Internet User Share USA Asia-Pacific Internet Users (millions) Wireless Internet User Share W. Europe Internet Users (millions) Wireless Internet User Share Source: September 3, 2004 - Computer Industry Almanac 1 Motivation Trend 2- MIP connect the world Enterprise WLAN Home AAA Server Gateway, HA, FA GSM Corporate LAN Ethernet VPN Firewall 802.11 Access Points CDMA IP Backbone/ Mobile Internet IP EDGE WCDMA Public WLAN CDMA2000 Ethernet 802.11 Access Points Mobile devices can connect to office networks anytime from anywhere….through Multi-mode terminal w/MobileIP client & IPSec Client 1 Motivation Trend 3- Moving speed accelerates wealth accumulation High Speed Trains German maglev (magnetic levitation): 400 kilometers per hour(248mph) France's Lignes a Grande Vitesse (LGV): commercial speed of 320km/h(198mph) Japanese Maglev Test speed 581 km/h (361 mph) Shanghai China's maglev train: 429km/h (267mph ) 1 Motivation Trend 4- small cell to keep a high throughput 802.11b data rate 1mbps 2mbps 5.5mbps 11mbps Office distance 50m 40m 35m 25m 400m 270 160m Outdoors distance 550m Small Wireless Cells Advantage: •Higher throughput •More number of users •Diameter of cell decreases by N number of cells in a certain area increases by N² number of available channels increases by N² Disadvantage: •Mobile host crosses cells more often 11mbps—160m 1mbps—550m More number of handoffs Data rate vs distance 36mbps—20m 6mbps—100m 1 Motivation Mobility Enhanced Data-rate for GSM Evolution 3G, Wide Area Network(WAN) coverage Wide band Code Division Multiple Access 2G Code Division Multiple Access 2000 WPAN: Bluetooth(802.15): Rapid Mobility? 0.7-2 Mb/s data ra Vehicle tes, Cable replacement, Short distances, up to 10 m. WLAN: 802.11 Outdoors Walk Could Mobile IP combine all them together GSM Fixed and fulfill its job at high speed IS-95 WLAN IS-136 802.11 n Walk Indoors Office WPAN 0.1 1 Mobility vs data rate 10 100 Mbps Techniques to Reduce IEEE 802.11 Handoff latency MIP Mishra in [Mish03]: L2 handoff latencies discovery 90% and reauthentication Hierarchical MIP 10%. 802.11, 11a, 11b, 11g Cellular standardsIP In [Mish04], Mishra proposed to reduce reauthentication byofIAPP. Define the MAC andlatency PHY layer OSI 2. Related works HAWAII Handoff management frame Infrastructure, ad hoc [Shin04], propose selective scanning algorithms to Handoff procedure 2.1 Network Layer Handoff Management r 2.2 Wireless LAN Techniques to Reduce IEEE 802.11 Handoff latency:. These are L2 only algorithms 2.3 Wireless LAN Handoff Management 2.4 Low Latency Handoff Mechanisms for MIP over 802.11 Network 2.5inLocation Tracking Malki [Malk02] proposed two mobility protocols, pre- and post-registration, usi Kim in [Kim04] and Shin in educe discovery latency. ….. ng L2 trigger. In pre-registration, MN may communicate both oFA and nFA. In post-registration, data cached in nFA before the registration completed. GPS or signal These are cross layer strength solutionsto triangulate the location of MN 2.1 Network Layer Handoff Management Macro and Micro Mobility Home Agent Macro mobility Mobile IP Home Network INTERNET GFA Macro-mobility handoff Micro-mobility handoff Gateway FA HMIP Micro mobility CIP domain HAWAII Handovers in Micro Mobility transparent outside the domain 2.1 Network Layer Handoff Management Mobile IP –Macro-mobility management Specified in RFC3344- C. Perkins, Nokia Research Center, August 2002 Foreign Network Correspondent Node (CN) Foreign Agent (FA) Advertisement (FA,COA) Solicitation Register Packets sent by MN go Register (HA) directly to CN Mobile Node(MN) Home Agent (HA) Home Network Mobile Node • When mobile node (MN) moves to a foreign network it obtains a care-of-address (COA) from the foreign agent (FA) that registers - Triangle Routing it with the home agent (HA) • Each mobile node has a home network, home address and - Reverse Tunnel • COAhome is used by HA to tunnel packets to MN agent 2.1 Network Layer Handoff Management Hierarchical Mobile IP –Micro-mobility CN Binding • Tree-like structure of FAs • Hierarchical Tunneling • Registration is Regionalized inside a domain To eliminate Handoff Traffic •Regional Registration Request & Reply HA MNFA1 INTERNET MNFA2 MNFA4 FA4 MNCOA MH FA1 MNFA3 MNFA5 FA2 FA5 MNCOA MH FA3 FA6 MNCOA MH MNFA6 FA7 Problems addressed in this dissertation • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: • Where does the latency come from ? • How much ? • What should we do with them ? • A deep examination of the handoff latency • Designing and Implementing high performance and scalability solution or protocols for rapid mobility Problems addressed in this dissertation • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: • A deep examination of the handoff latency • Where does the latency come from ? • How much ? • What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility 3. Performance at different speeds Experimental Scenario: In this scenario, a rapid moving MN will travel trough 8 APs. Each AP is wired to a FA. The distance between every two consecutive APs is d= 250m, 500m or 1000m. The moving speed of MN is V, varying from 10m/s to 80m/s. CN interne t HA FA1 AP1 FA2 AP2 FA3 AP3 FA4 AP4 FA5 AP5 FA6 AP6 FA7 AP7 FA8 AP8 3. Performance at different speeds RAMON: a Rapid Mobile Network emulator Originally developed by Edwin A. Hernandez in 2002. Rebuilt and extended by Jun Tian since 2003. Attenuators Attenuators are are pro-gr progr am controllable 3. Performance at different speeds •Control A ttenuators Antenna1 FA1 HUB1 192.168.1.1 AP1 Attenuator1 192.168.1.3 Antenna2 Controller FA2 HUB2 192.168.2.1 AP2 Attenuator2 COM 192.168.2.3 Antenna3 FA3 192.168.3.1 HUB3 AP3 Attenuator3 192.168.3.3 Emulator 192.168.1.2 MN 192.168.2.2 192.168.4.5 192.168.3.2 By increasing or decreasing the signal strength of one AP, we can e 192.168.4.2 192.168.4.1 Internet mulate the MN moving towards or HA away from the AP. By varying th COM 10.3.3.14 e increasing or decreasing speed of the signal strength, we can emul •manipulates the Attenuators ate the speed change of the MN. •A router Architecture of RAMON •Run emulation 3. Performance at different speeds time-sequence graph throughput graph Speed 20 m/s distance 1000m 3. Performance at different speeds time-sequence graph throughput graph Speed 80 m/s distance 1000m 3. Performance at different speeds time-sequence graph throughput graph Speed10 distance 500m 3. Performance at different speeds time-sequence graph throughput graph Speed40 distance 500m 3. Performance at different speeds Table 1: Average Throughput at Different Speeds and AP Distances. Speed (m/s) AP distance (m) Bytes transferr ed (kB) Travel Time (s) Average throughp ut (kB/s) Total handoff time(s) Effective time(s) PMaxavg (kB/s) Handoff Rate (FAs/s) 20 1000 78000 396 196.970 58 338 232.5 0.02 40 1000 33000 197 167.512 57 140 234.31 0.04 60 1000 16700 130.5 127.969 56 74.5 234.07 0.06 80 1000 9200 98.5 94.359 57 41.5 232.673 0.08 10 500 78500 397 197.733 58 339 233.01 0.02 20 500 33100 198 167.172 56 142 234.4 0.04 30 500 16600 129 128.682 56 73 232.86 0.06 40 500 9200 98 93.877 58 40 232.8 0.08 3. Performance at different speeds Conclusion 1: •The total handoff time doesn’t change with speed Conclusion 2 : •Effective time/total travel time ratio drops when speed goes up •This is the reason why high speed has low throughput 3. Performance at different speeds Average Throughputs vs Speeds. 3. Performance at different speeds Analysis: Let Pavg--Average throughput PMaxavg – Average throughput without handoff Ttravel– Total travel time Teffective – Total effective time for ftp transmission Thandoff- Total handoff time while traveling Khandoff – The number of handoffs while traveling thandoff – Average handoff time among 7 times of handoff Pavg = (Pmaxavg / Ttravel ) x Teffetive = Pmaxavg (Ttravel – Thandoff ) / Ttravel = Pmaxavg (1 – Thandoff / Ttravel) = Pmaxavg( 1 – Khandoff x thandoff / Ttravle) = Pmaxavg( 1 – (Khandoff / Ttravle ) x thandoff )) 3. Performance at different speeds Conclusion 3: Above analysis Pavg = Pmaxavg( 1 – (Khandoff / Ttravle ) x thandoff )) Conclusion 1 thandoff doesn’t change The change of Pavg is caused by Khandoff / Ttravel ratio. The performance of MIP over WLAN is related to the ratio of the number of handoffs/ total travel time, which is the MN handoff rate rh . rh = v/d. The ratio of speed and cell size(AP distance). This is different from what was previously believed. Kbytes/sec 3. Performance at different handoff rate 200 180 160 140 120 100 80 0 0.02 0.04 0.06 0.08 Handoff rate FA/s At handoff rate 0.02 FAs/s, the average throughput is 197.35 kB/s . When the handoff rate goes up to 0.08 FAs/s, the average throughput drops to 94.118 kB/s Problems addressed in this dissertation • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? Answer: • A global view of the handoff latency – Breakdown the handoff latency to see: Pavg• = Pmaxavg ( 1 – ofrhthex handoff t )) A deep examination latency handoff equation 2 • Where does the latency come from ? • How much ? • What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility Problems addressed in this dissertation • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: • A deep examination of the handoff latency • Where does the latency come from ? • How much ? • What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility 4. MIP over WLAN handoff latency CN 1 MN moves, signal decay or signal interruption causes MN probe request to all channels, available AP’s probe response to MN, MN selects best nAP and sends Authentication request and after gets Authentication ACK sends Reassociation Request and gets Reassociation Response. L2 handoff finished. HA GFA FA 1 AP1 FA 2 2. AP2 M N MN finds an nFA on its local network by the Agent Discovery process. After received 3 time of Agent Advertisement from nFA, MN sends Registration Request. nFA forward it to HA or GFA. HA or GFA reply with Registration Reply. L3 handoff finished. 3 Then MN begin to recover interrupted communi cation. L4 latency Global view of MIP/WLAN handoff thandoff = tL2handoff + tL3handoff + tL4handoff (equation 1) Problems addressed in this research • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: Answer: • A deep examination of the handoff latency • Where does the latency come from ? thandoff• How = tL2handoff + t L3handoff + tL4handoff (equation 1) much ? • What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency • Studying the effect ofexamination Rapid Mobility the performance of Mobile A deep of the on handoff latency Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: •Where does the latency come from ? •How much ? •What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency Data package L3 HO signal L2 HO signal MN oAP nAP oFA HA CN nFA L4 TCP retransmission belay:TCP TheACK retransmission timer at the sender is doubled with ea L2 movement detection ch unsuccessful retransmission in order to reduce the retransmission rate. Thus Probeattempt, request when theAPMN is reconnected, TCP will take a long time to recover from such a reductio L2 searching L2 Probe response n and data will not be transmitted for a period of time. ( TCP exponential backoff retran Authentication smission policy) L2 reassociation Reassociation request Reassociation response L2 movement detection delay: The strength of received signal degrades below a certain threshold, be caused by collision, radio signal fading, or AP is out of range. The MIPthis agentmay discovery MIP Advertisement STA(MN) detect the lack of radio connectivity based on weak received signal reported by L2L3searching delay: If transmission remains unsuccessful, the STA start scanning each Request During this period, the TCP ACKs sent by the physical layer or failed frame Registration transmissions. MIP registrationMN sends L2 reassociation: reassociation request to selected andrespones. waits for reassoci channel by bradcasting a probe-request frame and waiting for AP probe Registration Reply MN to CN keeping lost, as well as the TCP packets from oFA to MN. ation response L3 agent discovery delay: MN got agent advertisement from nFA through nAP. If MN isL4usingTCP active mode, it will send agent solicitation ask for agent advertisement. retransmission L3 registration delay: MN sends registration request to HA or GFA and waits for reply. • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – Breakdown the handoff latency to see: •Where does the latency come from ? Answer t L2handoff •How much ? =t +t +t •What should we do withL2reassociation them ? L2seraching L2detection (equation 3) t = t + t (equation 4) • Designing and Implementing high performance and L3handoff mipagentdicovery mipregistration mobility t scalability = t solution or protocols for rapid (equation 5) L4handoff t handoff tcp-back-off =t t L2detection tcp-back-off +t L2seraching +t L2reassociation +t mipagentdicovery +t mipregistration ( equation 6) + 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency • Studying the effect of Rapid Mobility on the performance of Mobile Networking Protocol. In particular, we examine the effects of: – How speed affects the performance ? • Performance/speed relationship – What constructs the handoff latency ? • A global view of the handoff latency – Breakdown the handoff latency to see: •Where does the latency come from ? •How much are the latencies ? •What should we do with them ? • Designing and Implementing high performance and scalability solution or protocols for rapid mobility 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency Experimental Scenario: In this scenario, a rapid moving MN will travel trough 8 APs. Each AP is wired to a FA. The distance between every two consecutive APs is d= 250m, 500m or 1000m. The moving speed of MN is V, varying from 10m/s to 80m/s. Pick out 20 experiments data. CN interne t HA FA1 AP1 FA2 AP2 FA3 AP3 FA4 AP4 FA5 AP5 FA6 AP6 FA7 AP7 FA8 AP8 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency exp# 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 avg avg Delay L2 movement detection 1.033 1.064 1.133 1.032 1.044 1.131 1.009 1.120 1.023 1.039 1.100 1.013 1.021 1.006 1.104 1.003 1.110 1.100 1.302 1.098 1.074 L2 AP searching 0.061 0.044 0.063 0.100 0.065 0.057 0.056 0.060 0.059 0.076 0.045 0.049 0.051 0.043 0.069 0.064 0.054 0.064 0.056 0.044 0.059 1.143 L2 reassociation 0.005 0.009 0.006 0.008 0.003 0.004 0.010 0.006 0.026 0.005 0.030 0.010 0.009 0.017 0.006 0.013 0.010 0.006 0.009 0.004 0.010 MIP agent discovery 2.996 1.945 3.023 2.563 2.756 2.578 2.436 3.001 2.213 3.008 2.770 2.545 3.001 2.600 2.598 2.674 2.783 3.012 2.349 2.404 2.660 2.746 MIP registration 0.073 0.042 0.052 0.050 0.052 0.043 0.060 0.704 0.054 0.053 0.041 0.042 0.065 0.046 0.047 0.062 0.054 0.057 0.070 0.062 0.086 TCP back_off 5.058 6.01 5.345 5.323 5.125 5.004 5.625 5.002 4.998 5.006 5.728 4.768 5.202 5.312 4.544 4.806 5.705 5.602 5.71 5.172 5.253 5.253 Handoff Latency 9.226 9.511 9.622 9.076 9.045 8.817 9.196 9.893 8.373 9.187 9.714 8.427 8.896 9.024 8.368 8.622 9.716 9.841 9.496 8.784 9.142 9.142 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency Data package L3 HO signal L2 HO signal MN oAP nAP oFA HA CN nFA TCP ACK 1.074 L2 movement detection L2 delay 1.143 0.059 L2 AP searching 0.010 L2 reassociation Conclusions: L3 delay 2.746 L4 delay 5.253 Probe request Probe response Reassociation request Authentication Reassociation response MIP Advertisement •The largest latency is due to TCP, then Layer 3. 2.660 MIP agent discovery Registration Request Registration Reply •If0.086 we MIP reduce registrationthe L2 and L3 delay, L4 delay will be reduced exponentially. 5.253 TCP retransmission •This dissertation focuses on L3 latency. Handoff delay : 9.142 5. Speed Adaptive Mobile IP Pavg = Pmaxavg( 1 – rh x t handoff )) (equation 2) Handoff Rate rh = v/d MN moves through how many APs per second. rh = Khandoff / Ttravel (equation 7) Where Khandoff is the number of handoffs occurred during the MN traveling. Ttotal is MN’s total travel time. Lower handoff rate has higher throughput To reduce rh without changing total travel time, have to decrease the number of handoffs. The optimal is Khandoff = 0 5. Speed Adaptive Mobile IP Let N be total FA numbers on the way MN traveling. Let’s assume somehow M is the number of FAs MN can communicate with without L3 handoff delay, which means M is the number of FAs MN has registered to at that moment. The optimal is let M = N. But this costs too much resources, especially when the number of active MNs is large. Also we don’t know how long will the MN travel at the beginning. 5. Speed Adaptive Mobile IP We call M the FA set size that MN registered. Questions: 1. How to decide FA set size M 2. How to guarantee MN can communicate with a FA set almost like to do with a single FA. For Q1 M thandoff rh 1 (equation 8) where thandoff is the handoff time, rh is the handoff rate. Here we use the experimental average handoff time 9.142s for thandoff. rh is dynamic. For example, at speed 40m/s, AP distance 500m, M = | 9.142 x 40/500 | + 1 = 2. At speed 80m/s, AP distance 500m, M = 3. 5. Speed Adaptive Mobile IP For Q2, How to guarantee MN can communicate with a FA set almost like to do with a single FA. Current FA pre-registers MN with M potential FAs to reduce L3 handoff latency, at the same time let IP packets be multicast to those M FAs in this FA set. So MN won’t feel any handoff delay at the IP level. By above 2 steps, the set of FAs that MN can talk to without L3 latency was extended from one point in stationary state, to a line at high speed. M= 1 rh = 0 M=2 0<rh < 0.109 M=3 0.109<rh< 0.218 FA Set Size vs Handoff Rate M=4 0.218<rh< 0.328 5. Speed Adaptive Mobile IP Speed extension MN’s registration message is extended by speed extension. According to Mobile IP Vendor/Organization-Specific Extensions[RFC3115]. Two Vendor/Organization Specific Extensions are allowed for MIP, Critical (CVSE) and Normal (NVSE) Vendor/Organization Specific Extensions. The basic difference is when the CVSE is encountered but not recognized, the message containing the extension must be silently discarded, whereas when a NVSE is encountered but not recognized, the extension should be ignored, but the rest of the Extensions and message data must still be processed. We use the NVSE extension. 5. Speed Adaptive Mobile IP Normal Vendor/Organization Specific Extension Type = 134 for NVSE extension. Length is the size in bytes of the extension, not including the type a nd length bytes. Verdor/org-ID is assigned in RFC 1700. set 5205. Vendor-NVSE-Type Indicates the particular type of Vendor-NVSEExtension. Set as 12 for speed extension. Vendor-NVSE-Value: the value of MN moving speed. 5. Speed Adaptive Mobile IP interne t CN HA 5 3 6 7 13 FA1 FA2 4 FA3 FA4 12 AP1 AP2 AP3 AP4 1 2 8 9 10 11 MN the GFA received these it FA builds up tunnels downwards to re ea Whenever the MN setups the Link-layer contact with the FA, the later forwards Whenever The firstsame FA the relays MNor needs the registration to handoff request to aregistration new torequest FA upper set,requests, FA after or itnew HA(step gets that 3).many timesthe of registration agent adverti AtWhen the time, itHA still sends registration to the set with up-to-date speed informa ch FAother and registration reply(step 6 and 7). When FA received ply toMN the former(step 8, with 9the orregistration 10). These FAs relay request to upper or HA the as well(step 5). sements which is care-of-address determined by speed(step 1), tion (step 11).responses The very first FA in thisagent set decupsulate theFA message and set up new FAregistratio set. Forwar The gets the from advertisement message(step 10 or a9) orthe registration repl Meanwhile, it decapsulates the speed extension, refill the MIP header and authentication extension n reply, it builds upwards the orprocess. HA. registration request.(12, and repeat theGFA above ydmessage(step 9 or up 10),tunnel and13) begins datatocommunication. it sends and thena forward registration it torequest other FAs(M-1 with up-to-date FAs) in this moving FA set(step speed information 4). to the very first FA in a new FA set (step 2). This FA will calculate the handoff rate and M 5. Speed Adaptive Mobile IP Same scenario as above except SA-MIP is installed. Time-sequence graph at speed 60m/s and AP distance 1000m 5. Speed Adaptive Mobile IP Same scenario as above except SA-MIP is installed. Time-sequence graph at speed 80m/s and AP distance 1000m 5. Speed Adaptive Mobile IP Average throughput at different speeds and AP distances. Speed (m/s) AP distance (m) Bytes transferred (kB) Travel Time(s) Arg throughput (kB/s) Handoff Rate (FAs/s) 20 1000 85000 399 213.03 0.02 40 1000 37500 198 189.39 0.04 60 1000 19400 130 149.23 0.06 80 1000 11600 99 117.17 0.08 10 500 84400 398 212.06 0.02 20 500 37400 198 188.89 0.04 30 500 19500 131 148.55 0.06 40 500 11500 98 117.34 0.08 Kbytes/sec 5. Speed Adaptive Mobile IP 220 200 SA -MIP 180 MIP 160 140 (212.55 - 197.35) /197.35 = 7.69% 120 (189.14 - 167.34) /167.34 = 13.02% (148.89 - 128.32) /128.32 = 15.97% 100 (117.25 - 94.12) /94.12 = 24.58% 80 0 0.02 0.04 0.06 0.08 Handoff rate FA/s Average Throughput of Speed-Adaptive MIP 6. Summery and Future Works Contributions: Evaluate the rapid mobility of MIP over wireless LAN in a laboratory environments. Depicted the relationship between the performance and the moving speed of MN Quantitatively analyzed the handoff latencies of the MIP over wireless LAN Speed Adaptive MIP is proposed and evaluated Future Works Speed adaptive scheme should be applied to layer 2 and layer 4 handoff latencies. Publications: 1. J. Tian, A. Helal, "Rapid Mobility of MIP over WLAN," International Co nference on Computer Networks and Mobile Computing (ICCNMC'05), Zh angjiajie, China, Aug 2-4, 2005. Lecture Notes in Computer Science, Sprin ger-Verlag GmbH, ISSN: 0302-9743, ISBN: 3-540-28102-9, Volume 3619/ 2005 2. Jun Tian and A. Helal, "Performance of MIP over WLAN in Rapid Movi ng Environments," The 4th ACS/IEEE International Conference on Comput er Systems and Applications, Dubai/Sharjah, UAE, March 8-11, 2006 3. J. Tian and A. Helal, "Speed Adaptive MIP" submitted to the IEEE Wirel ess Communication&Network Conference, to be held in Las Vegas, NV, Ap ril 2006 4. J. Tian and A. Helal, "Speed Adaptive MIP over Wireless LAN," Submitt ed to Journal of Wireless Communications and Mobile Computing • backups 1.1 Network Layer Handoff Management Cellular IP –Micro-mobility(Campbell ) Mobile IP enabled Network HA •Inside the CIP network, Uplink data GW Cellular IP network BS1 packets are routed from MN to the gateway on a hop-by-hop basis. •The path taken by these packets is cached in route cache of base stations. •CIP uses mobile originated data packets to maintain reverse path. This path is used to route downlink packets addressed to a mobile host. IP router CIP Node CIP BS 10Mb 2ms BS2 CIP routing BS4 802.11 MN IP routing IP tunnelling BS3 1.1 Network Layer Handoff Management Cellular IP –Micro-mobility Routing and Paging Update Mobile IP enabled Network HA Routing update packet Paging update packet GW Cellular IP network Idle MN Active MN 1.1 Network Layer Handoff Management Cellular IP –Micro-mobility Handoff MN listen/send to only one BS(TDMA network ) Hard Handoff packet loss cannot be eliminated MN listen/send to two or more BS simultaneously (CDMA network) Semi-Soft Handoff good performance eliminates packet loss 3. Performance at different speeds Rebuilding of RAMON: • Update hardwire: 3 FA, MN, new CISCO 350 APs, control board (Tarek Kaissi) •Re-construct the testbed: NAT on HA, routing table on Emulator(h elp from Edwin) •Software installation and configuration: •Linux kernel 2.4.20, •HUT dynamic MIP implementation version 0.8.1 (Helsinki Un iversity of Technology) •Erase architecture delay 4. Quantitative Analysis of the MIP over Wireless LAN Handoff Latency A deep examination of the handoff latency L2 movement detection delay: The strength of received signal degrades below a certain threshold, this may be caused by collision, radio signal fading, or AP is out of range. The STA(MN) detect the lack of radio connectivity based on weak received signal reported by the physical layer or failed frame transmissions. During the period, the TCP ACK sent by MN to CN keeps lost. L2 searching delay: If transmission remains unsuccessful, the STA start scanning each channel by bradcasting a probe-request frame and waiting for probe respones. L2 reassociation: MN sends reassociation request to selected AP and waits for reassociation response. tL2handoff = tL2detection + tL2seraching + tL2reassociation (equation 4) L3 agent discovery delay: MN got agent advertisement from nFA through nAP. If MN is using active mode, it will send agent solicitation ask for agent advertisement. L3 registration delay: MN sends registration request to HA or GFA and waits for reply. tL3handoff = tmipagentdicovery + tmipregistration (equation 4) L4 TCP retransmission belay: The retransmission timer at the sender is doubled with each unsuccessful retransmission attempt, in order to reduce the retransmission rate. Thus when the MN is reconnected, TCP will take a long time to recover from such a reduction and data will not be transmitted for a period of time. ( TCP exponential backoff retransmission policy) tL4handoff = ttcp-back-off (equation 5) thandoff = tL2detection + tL2seraching + tL2reassociation + tmipagentdicovery + tmipregistration + ttcp-back-off ( equation 6) 5. Speed Adaptive Mobile IP Whenever the MN needs to handoff to a new FA set, after it gets that many times of agent advertise ments which is determined by speed(step 1), CN it sends a registration request with up-to-date handoff HA w FA set (step 2). interne t rate information to the very first FA in a ne The first FA relays the registration request to upper FA or HA(step 3). 5 3 Meanwhile, it decapsulates the 6speed extension, refill the MIP header and authentication extension 7 13 and then forward it to other FAs(M-1 FAs) in this FA set(step 4). FA1 4 FA2 FA3 FA4 12 These other FAs relay the registration request to upper FA or HA as well, just like the request com es from the MN(step 5). When the GFA or HA received these registration requests, it builds up tunnels downwards to each AP1 AP2 AP3 AP4 FA and responses with registration reply(step 6 and 7). When the FA received the registration reply, it builds up tunnel upwards to the GFA or HA. When ever the1MN setups the Link-layer contact with the FA, the later forwards the registration reply to 2 8 the former(step8, 9 or 10). 9 10 11 The MN gets the care-of-address from agent advertisement message(step 10 or 9) or registration re MN ply message(step 9 or 10), and begins data communication. At the same time, it sends registration requests to the new FA with up-to-date speed information (s tep 11). This new FA decapsulates the message, sets up a new FA set, forwards the request(12,13) and repeats the above process.