10.07.2015 Views

Final Exam: Mobile Networking (Part II of the course “Réseaux et ...

Final Exam: Mobile Networking (Part II of the course “Réseaux et ...

Final Exam: Mobile Networking (Part II of the course “Réseaux et ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Final</strong> <strong>Exam</strong>: <strong>Mobile</strong> <strong>N<strong>et</strong>working</strong>(<strong>Part</strong> <strong>II</strong> <strong>of</strong> <strong>the</strong> <strong>course</strong> “Réseaux <strong>et</strong> mobilité”)Pr<strong>of</strong>. J.-P. HubauxFebruary 12, 2004Duration: 2 hours, all documents allowedPlease write your answers on <strong>the</strong>se she<strong>et</strong>s, at <strong>the</strong> end <strong>of</strong> each question;use extra she<strong>et</strong>s if necessary (put your name on <strong>the</strong>m)You may write your answers in English or in French.The total number <strong>of</strong> points (65) corresponding to <strong>the</strong> proposed questions is higher than <strong>the</strong> number <strong>of</strong>points (50) required to obtain <strong>the</strong> highest mark.L<strong>et</strong> p be <strong>the</strong> number <strong>of</strong> points you obtain; your final mark to this exam will be min(6, (1 + p/10)).This document contains 13 pages.Student First name:Last name:(answers to <strong>the</strong> questions are shown in italic and blue)


1 WLANs (30 points)1.1 IEEE 802.11 (10 points)Consider <strong>the</strong> following scenario:S1 D S2S1 and S2 send CBR / UDP traffic to <strong>the</strong> common destination D, using IEEE 802.11 (DCF).1.1.1 Scenario 1: Consider S1, S2 and D all within receive range <strong>of</strong> each o<strong>the</strong>ra) When <strong>the</strong> basic scheme is used (no RTS/CTS):• Describe a collision (what happens before, during, and after)S1DDIFSDataSIFSACKDIFSDatabkf = rand(doubled CW)DataSIFSACKTimeS2Data• What does <strong>the</strong> collision probability depend on ?Contention window sizePack<strong>et</strong> rate (but not pack<strong>et</strong> size)Number <strong>of</strong> stations• Consider <strong>the</strong> total channel capacity to be X b/s. S1 and S2 send data flows at rates Y 1 = Y 2 >X. How is <strong>the</strong> channel capacity shared b<strong>et</strong>ween S1 and S2?S 1 → X/2S 2 → X/22


) When RTS/CTS is used• What are <strong>the</strong> changes to <strong>the</strong> previous answers ? Comment also on <strong>the</strong> throughput and fairness.Channel is still equally sharedwith more overhead (throughput decreases)Shorter collision duration1.1.2 Scenario 2: Consider that S1 and S2 are hidden from each o<strong>the</strong>r, only D hears both (S1cannot even sense S2, and vice versa)a) When <strong>the</strong> basic scheme is used (no RTS/CTS):• What are <strong>the</strong> changes to <strong>the</strong> previous answers ?Collision probability depends also on <strong>the</strong> pack<strong>et</strong> sizeThroughput decreases considerablythroughput S1 = throughput S2 = X ′ /2 X ′ /23


1.2 UDP and TCP over IEEE 802.11 (13 points)Consider <strong>the</strong> following scenario:S0192.168.10.110Mb/sAP192.168.10.2192.168.12.1MS1192.168.12.22Mb/sMS2192.168.12.3where S0 is a station wire-connected to <strong>the</strong> access point (AP). MS1 and MS2 are wireless nodesassociated to <strong>the</strong> AP, using IEEE 802.11 (DCF basic mode, i.e. no RTS/CTS). The cable capacity is10 Mb/s, <strong>the</strong> total wireless channel capacity is 2 Mb/s. Note that all downlink flows share <strong>the</strong> samequeue at <strong>the</strong> AP.Scenario 1: Consider <strong>the</strong> channel to be clear (no noise). S0 sends 1 CBR/UDP flow to MS1at 5 Mb/s, and 1 CBR/UDP flow to MS2 at 5 Mb/s. (Both flows go through <strong>the</strong> AP)• What is <strong>the</strong> throughput and pack<strong>et</strong> drop rate for each flow ?Throughput = 1 Mbps, drop rate = 4 Mbps(or, more precisely) Throughput = 0.8 Mbps, drop rate = 4.2 Mbps• What is <strong>the</strong> probability <strong>of</strong> a collision on <strong>the</strong> radio channel ? Explain why.P(collision) = 0, i.e. no collisions, since only <strong>the</strong> AP is transmitting4


Scenario 2: S0 sends 1 CBR/UDP flow to MS1 at 5 Mb/s, and 1 CBR/UDP flow to MS2 at 5Mb/s, (both flows go through <strong>the</strong> AP). The channel b<strong>et</strong>ween <strong>the</strong> AP and MS2 has a high errorrate.• Comment on <strong>the</strong> throughputs received at MS1 and MS2. Compare <strong>the</strong>m to <strong>the</strong> result <strong>of</strong> Scenario1.T hroughput S2


• Compare <strong>the</strong> throughput to that <strong>of</strong> Scenario 2.T hroughput < 2Mbps, because <strong>of</strong> <strong>the</strong> overheadand because <strong>of</strong> <strong>the</strong> collisions b<strong>et</strong>ween <strong>the</strong> AP (data) and MS1 (ACKs)• What is <strong>the</strong> pack<strong>et</strong> drop rate ?≈ 0, because TCP adapts to <strong>the</strong> available data rateScenario 4: The channel is clear, S0 sends 1 TCP flow to MS1 and 1 TCP flow to MS2, (bothflows go through <strong>the</strong> AP).• Describe a collision (on <strong>the</strong> wireless channel) in this case, followed by a successful transmission,using <strong>the</strong> following figure:TimeS0APDataDataDataSIFSDIFSDatabkf = rand(doubled CW)DataMS1ACKTCP−ACKMS2Scenario 5: Since TCP (at S0) is unable to distinguish if a pack<strong>et</strong> was dropped because <strong>of</strong>congestion at <strong>the</strong> AP or on <strong>the</strong> wireless channel, one possible solution is to install a “TCP proxy”at <strong>the</strong> AP, that sends back TCP-ACKs whenever a pack<strong>et</strong> is successfully received from <strong>the</strong> wiredpart, and forwards <strong>the</strong> pack<strong>et</strong> on <strong>the</strong> wireless channel.• What are <strong>the</strong> advantages ?TCP will not r<strong>et</strong>ransmit in case <strong>of</strong> channel pack<strong>et</strong> loss, MAC will do it faster than TCP woulddo.TCP does not perform congestion control when loss is on <strong>the</strong> channel (not necessarily congestion).6


• What are <strong>the</strong> drawbacks ?TCP will adapt to <strong>the</strong> wired link capacity, which has a higher data rate than what can behandled at <strong>the</strong> wireless channel (which needs an infinite queue length)1.3 On <strong>the</strong> vulnerabilities <strong>of</strong> IEEE 802.11 (7 points)• Why is SIFS shorter than DIFS ?To give priority to ACKs, which must be transmitted before any new data pack<strong>et</strong>.To increase his share <strong>of</strong> <strong>the</strong> throughput, a cheater may use <strong>the</strong> following cheating m<strong>et</strong>hodsin IEEE 802.11:• The cheater sends his frame before <strong>the</strong> end <strong>of</strong> DIFS• The cheater uses contention windows smaller than those specified in <strong>the</strong> standardQuestions:• Which one is more efficient ? Why ?DIFS. It gives <strong>the</strong> cheater absolute priority (like <strong>the</strong> priority SIFS gives to ACKs)• Which one is easier to d<strong>et</strong>ect ? Why ?DIFS. Small contention windows give statistical transmission successes, while DIFS gives absolut<strong>et</strong>ransmission success.7


2 Cellular N<strong>et</strong>works (15 points)2.1 Cellular principles• What is <strong>the</strong> difference b<strong>et</strong>ween multiple access (e.g., FDMA) and duplex (e.g., FDD)? (2 points)– Multiple access allows a number <strong>of</strong> users to access a single radio-frequency (RF) channelwithout interference.– Duplex allows both ends <strong>of</strong> a communication channel to send and receive signals at <strong>the</strong>same time.• Is it possible to have s<strong>of</strong>t handover in GSM? Why? (2 points)No. Because in GSM, which uses TDMA instead <strong>of</strong> CDMA, a hands<strong>et</strong> cannot communicatewith more than one base station at <strong>the</strong> same time.• Explain in a few sentences what <strong>the</strong> near/far effect is and how it is coped with in cellular n<strong>et</strong>works.(2 points)Near/far effect is a problem specific to CDMA-based cellular n<strong>et</strong>works. If <strong>the</strong>re is more thanone active user, <strong>the</strong> interference power with respect to a given user (reference user) at <strong>the</strong> basestation is <strong>the</strong> transmitted power <strong>of</strong> o<strong>the</strong>r users (non-reference users) suppressed by a factordepending on <strong>the</strong> code used by <strong>the</strong> CDMA system. However, if some non-reference users arecloser to <strong>the</strong> base station than <strong>the</strong> reference user, it is possible that <strong>the</strong> interference caused by<strong>the</strong>se non-reference users (however suppressed) has more power than <strong>the</strong> reference user. As aresult, <strong>the</strong> SIR <strong>of</strong> <strong>the</strong> reference user becomes too lower to allow correct reception. Control <strong>the</strong>transmission power <strong>of</strong> each user such that <strong>the</strong> received power at <strong>the</strong> base station is equal forall user is a solution.• Consider a cellular n<strong>et</strong>work with a given propagation loss exponent α; describe two ways toachieve a certain minimum required SIR? (3 points)– D<strong>et</strong>ermine a co-channel reuse ratio Q = √ 3N. Note that a larger Q implies a largercluster size and less channels assigned to each cell.– Use sectoring technique to fur<strong>the</strong>r divide a cell into multiple sectors. This means that <strong>the</strong>usage <strong>of</strong> directional antennae and a sacrifice <strong>of</strong> trunk efficiency within a cell.8


2.2 Capacity <strong>of</strong> Cellular SystemsAssume that in a FDMA cellular system, a minimum SIR <strong>of</strong> 17dB is required for satisfactory downlinkchannel performance. Given a propagation loss exponent α = 4, propose three frequency reusepatterns for <strong>the</strong> system. Note that <strong>the</strong> cluster size N should be <strong>the</strong> least integer that satisfies <strong>the</strong> minimumrequired SIR; choosing an unreasonably large N will not be considered as a solution (Hint:partitioning a cell into three 120 o sectors is not <strong>the</strong> only feasible way to do sectoring). (6 points)S• No sectoringI = ( D R )4 1 6(i.e., <strong>the</strong>re are 6 co-channel interference cells).• 120 ◦ SsectoringI = ( D R )4 1 2(i.e., <strong>the</strong>re are 2 co-channel interference cells).• 60 ◦ SsectoringI = ( D R )4 (i.e., <strong>the</strong>re is only 1 co-channel interference cell).The requirement is that:S≥ 17dB = 50.12ISo, considering that ( D R )4 = 9N 2 and N can only take values <strong>of</strong> 1, 3, 4, 7, 9, · · ·, we have:⎧⎪⎨ 7 no sectoringN = 4 120 sectoring⎪⎩ 3 60 ◦ sectoring◦9


3 Ad hoc n<strong>et</strong>works (20 points)3.1 Throughput (6 points)In static multi-hop ad hoc n<strong>et</strong>works, <strong>the</strong> maximal throughput decreases approx. with √ N, where Nis <strong>the</strong> number <strong>of</strong> nodes in <strong>the</strong> n<strong>et</strong>work.a) Explain precisely which throughput decreases at that rate.The throughput that decreases at that rate is <strong>the</strong> throughput <strong>of</strong> <strong>the</strong> end-to-end traffic b<strong>et</strong>weenrandomly chosen sources and destinations.b) Explain <strong>the</strong> intuition behind this throughput decrease.In an ad hoc n<strong>et</strong>work, nodes that are sufficiently distant can transmit concurrently. Therefore <strong>the</strong> totalamount <strong>of</strong> data that can be simultaneously transmitted for one hop increases linearly with <strong>the</strong> totalarea <strong>of</strong> <strong>the</strong> ad hoc n<strong>et</strong>work. If node density is constant, this means that <strong>the</strong> total one-hop capacityis O(n), where n is <strong>the</strong> total number <strong>of</strong> nodes. However, as <strong>the</strong> n<strong>et</strong>work grows larger, <strong>the</strong> number<strong>of</strong> hops b<strong>et</strong>ween each source and destination may also grow larger, depending on communicationpatterns. One might expect <strong>the</strong> average path length to grow with <strong>the</strong> spatial diam<strong>et</strong>er <strong>of</strong> <strong>the</strong> n<strong>et</strong>work,equivalently <strong>the</strong> square root <strong>of</strong> <strong>the</strong> area, i.e., O( √ n). With this assumption, <strong>the</strong> total end-to-endcapacity is roughly O(n/ √ n), and <strong>the</strong> end-to-end throughput available to each node is O(1/ √ n).c) Explain how each <strong>of</strong> <strong>the</strong> following situations influences throughput:(i) The message sources and destinations are necessarily neighborsIf <strong>the</strong> message sources and destinations are neighbors, all <strong>the</strong> communications are single-hop and <strong>the</strong>throughput is high.(ii) The nodes send video traffic at a constant rate to randomly chosen destinationsIf destinations are chosen at random, <strong>the</strong> communications may be multi-hop, which decreases <strong>the</strong>end-to-end throughput. Fur<strong>the</strong>rmore, <strong>the</strong> video traffic assumes streaming, which induces more interference,more collisions and <strong>the</strong>refore low end-to-end throughput.10


(iii) The nodes communicate by occasionally sending messages (e.g., SMS) to randomlychosen destinations.As explained previously, multi-hop communications decrease end-to-end throughput. However, occasionaltraffic leads to less interference and collisions than in <strong>the</strong> previous case; <strong>the</strong> end-to-endthroughput is <strong>the</strong>refore higher than for <strong>the</strong> previous case.(iv) Nodes are mobile, but <strong>the</strong> application is not latency-sensitive.If <strong>the</strong> application is not latency-sensitive, this means that buffering and r<strong>et</strong>ransmissions are possible.Therefore, <strong>the</strong> throughput is expected to be high.3.2 Routing (6 points)Today’s mainstream proposals for ad hoc n<strong>et</strong>work routing are DSR and AODV protocols.a) Would it be efficient to make use <strong>of</strong> AODV or DSR protocols for routing in today’s Intern<strong>et</strong> ?If Yes, explain how. If No, explain why.No. Indeed, <strong>the</strong> Intern<strong>et</strong> is a wired n<strong>et</strong>work where <strong>the</strong> majority <strong>of</strong> <strong>the</strong> nodes are static and where <strong>the</strong>quantity <strong>of</strong> information exchanged is very important; a reactive routing protocol such as AODV andDSR is <strong>the</strong>refore not suitable for this kind <strong>of</strong> n<strong>et</strong>work.11


) Assume that you need to implement a messaging service in a mobile ad hoc n<strong>et</strong>work in whicheach user is sending messages at a rate <strong>of</strong> 1 message/10 minutes. Would you use AODV or DSRfor message routing? Explain why and how. If Yes, explain how. If No, what would be your o<strong>the</strong>rproposals?There is no need to establish a route b<strong>et</strong>ween <strong>the</strong> source and <strong>the</strong> destination if <strong>the</strong> message sendingfrequency is low (1 message/10 minutes). Indeed, both AODV and DSR flood <strong>the</strong> n<strong>et</strong>work with requestsin order to g<strong>et</strong> routes to a given destination. However, <strong>the</strong> r<strong>et</strong>urned route will be used to send onlyone pack<strong>et</strong> (<strong>the</strong> route will most likely be stale after 10 minutes), which makes no sense. An alternativesolution would be to flood <strong>the</strong> n<strong>et</strong>work with <strong>the</strong> message to be sent; <strong>the</strong> message will eventually reach<strong>the</strong> destination and <strong>the</strong> cost <strong>of</strong> <strong>the</strong> solution is lower than <strong>the</strong> cost <strong>of</strong> AODV or DSR.3.3 Security (8 points)a) How would you explain <strong>the</strong> fact that “Mobility breaks <strong>the</strong> security-routing dependency cycle”? Are<strong>the</strong>re o<strong>the</strong>r techniques by which <strong>the</strong> security-routing dependency cycle can be avoided?To establish a security association b<strong>et</strong>ween nodes A and B, we don’t rely on multi-hop routes (whichleads to <strong>the</strong> security-routing dependency cycle), we ra<strong>the</strong>r establish it when A and B are within <strong>the</strong>power range <strong>of</strong> each o<strong>the</strong>r; mobility increases <strong>the</strong> probability <strong>of</strong> encounter b<strong>et</strong>ween <strong>the</strong> two entities.O<strong>the</strong>r techniques can be used to avoid <strong>the</strong> security-routing dependency cycle (Preloading all pairs <strong>of</strong>keys into nodes, On-line au<strong>the</strong>ntication servers, ...).12


) Why is security important for routing? Describe one simple attack by which a single maliciousnode can disable all communication <strong>of</strong> its neighbors.If <strong>the</strong> routing protocol is not secure, an attacker can easily disrupt <strong>the</strong> communications in <strong>the</strong> ad hocn<strong>et</strong>work.E.g., by sending forged routing pack<strong>et</strong>s, an attacker could route all pack<strong>et</strong>s for some destination toitself and <strong>the</strong>n discard <strong>the</strong>m. This attack is called “Black hole attack”.c) Consider an ad hoc n<strong>et</strong>work running a DSR protocol. If all nodes in <strong>the</strong> n<strong>et</strong>work share pairwisekeys, and if all mutual communication b<strong>et</strong>ween nodes is encrypted with <strong>the</strong> keys that <strong>the</strong>y share (bothpack<strong>et</strong> headers and payload), does this make <strong>the</strong> n<strong>et</strong>work resistant to attacks? If yes, explain how. Ifno, describe <strong>the</strong> attacks that you believe can harm <strong>the</strong> n<strong>et</strong>work routing protocol operation.No. Indeed, even if all nodes in <strong>the</strong> n<strong>et</strong>work share pairwise keys, <strong>the</strong> “Black hole attack” describedabove is still possible.13

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!