15-441 Computer NetworkingNew Directions in DDoS Defense* (Peaceful Non-cooperation)Why it works better Slide 4OutlineAreasHow does XCP Work?Slide 8Slide 9Single TCP Flow Router with large enough buffers for full link utilizationExampleIf flows are synchronizedIf flows are not synchronizedTCP Modeling & TFRCSlide 17Slide 18Common Case in the Future InternetTrends: Density & ManagementTrends: Growing Application DiversityTrends: Spectrum ScarcityImplications: Spectrum ScarcityTrends: Growing Deployment DiversityTrends: New TechnologiesNew DirectionsSlide 27Slide 28Distributed Multiplayer Game ServerWhat do Multiplayer Games Look Like?DemoColyseus Architecture OverviewPlayer Update – SolutionMERCURY [Sigcomm 2004]Distributed Hash Tables (DHT)Using DHTs for Range QueriesSlide 37DHTs with Load BalancingSlide 39Ideal Link Structure15-441 Computer NetworkingLecture 27 – Research Directions12-05-06 Lecture 27: Research Directions 2New Directions in DDoS Defense*(Peaceful Non-cooperation)GuiltAttackerGood guyAttackWhat the !@#!@#$ ?I know you are attackingmachine A, here are the details of machine BYou have opened my eyes, I want to travel to the peak of Himalayas and meditate for the rest of my life*from Mohit Lad (UCLA)12-05-06 Lecture 27: Research Directions 3Why it works better FearAttackerGood guyAttackWhat the !@#!@#$ ?I know you are attackingmachine A, here are the details of machine BSeems like a trap, he must be a genius. Instead I am going to attack somebody willing to defend himself.12-05-06 Lecture 27: Research Directions 4Why it works better Lack of challengeAttackerGood guyAttackI am not going to say “What the !@#!@#$ ?” againI know you are attackingmachine A, here are the details of machine BMy fellow hackers will look down on me. Instead I am going to attack somebody willing to defend himself12-05-06 Lecture 27: Research Directions 5Outline•Transport•Wireless•Other stuff12-05-06 Lecture 27: Research Directions 6Areas•Router interactions•FQ, RED Blue, CHOKe, CSFQ, XCP…•Small buffer routers•New congestion control designs•Delay based (Vegas)•Long-term TCP fair (TFRC)•Others: bionomial, BIC/CUBIC•Other issues•Large bandwidth-delay product networks•Delay Tolerant Networks (DTN)•Congestion control outside TCP (congestion controlled UDP, general congestion management)12-05-06 Lecture 27: Research Directions 7Feedback Round Trip TimeCongestion WindowCongestion HeaderFeedback Round Trip TimeCongestion Window How does XCP Work?Feedback = + 0.1 packet12-05-06 Lecture 27: Research Directions 8Feedback = + 0.1 packet Round Trip TimeCongestion Window How does XCP Work?Feedback = - 0.3 packet12-05-06 Lecture 27: Research Directions 9 Congestion Window = Congestion Window + FeedbackRouters compute feedback without any per-flow state Routers compute feedback without any per-flow state How does XCP Work?XCP extends ECN and CSFQ12-05-06 Lecture 27: Research Directions 11Single TCP FlowRouter with large enough buffers for full link utilization12-05-06 Lecture 27: Research Directions 13Example•10Gb/s linecard•Requires 300Mbytes of buffering.•Read and write 40 byte packet every 32ns.•Memory technologies•DRAM: require 4 devices, but too slow. •SRAM: require 80 devices, 1kW, $2000.•Problem gets harder at 40Gb/s•Hence RLDRAM, FCRAM, etc.•Rule-of-thumb makes sense for one flow•Typical backbone link has > 20,000 flows•Does the rule-of-thumb still hold?12-05-06 Lecture 27: Research Directions 14If flows are synchronized•Aggregate window has same dynamics•Therefore buffer occupancy has same dynamics•Rule-of-thumb still holds.2maxWtmax2W�maxW�maxW12-05-06 Lecture 27: Research Directions 15If flows are not synchronizedProbabilityDistributionB0Buffer SizeW12-05-06 Lecture 27: Research Directions 16TCP Modeling & TFRC•Given the congestion behavior of TCP can we predict what type of performance we should get?•What are the important factors•Loss rate: Affects how often window is reduced•RTT: Affects increase rate and relates BW to window•RTO: Affects performance during loss recovery•MSS: Affects increase rate•TFRC – measure each of the above and set the transmit rate to that value32 pRTTMSSBW12-05-06 Lecture 27: Research Directions 17Outline•Transport•Wireless•Other stuff12-05-06 Lecture 27: Research Directions 18Areas•Ad hoc networks Mesh•Sensor networks scale•TCP over wireless DTNs & third world issues•Mesh networks•MAC protocols•Software defined radios•Chaotic wireless•Managing wireless networks12-05-06 Lecture 27: Research Directions 19Common Case in the Future Internet•Historic shift from PC’s to mobile computing and embedded devices…•>2B cell phones vs. 500M Internet-connected PC’s in 2005•>400M cell phones with Internet capability, rising rapidly•Sensor deployment just starting, but some estimates ~5-10B units by 2015INTERNETINTERNETWirelessEdge NetworkWirelessEdge NetworkINTERNETINTERNET~500M server/PC’s, ~100M laptops/PDA’s~750M servers/PC’s, >1B laptops, PDA’s, cell phones, sensors20052010WirelessEdge NetworkWirelessEdge Network12-05-06 Lecture 27: Research Directions 20Trends: Density & Management•Densities of unlicensed devices already high•Limits performance due to scarce spectrum•Need more spectrum or more efficient use•Channel allocation suggests poor management•Makes problems such as interference worse•Security management probably worse•Won’t get better need automation6 5111 211 1410 4Channel %Portland 8683 54San Diego 7934 76San Fran 3037 85Boston 2551 39#APs Maxdegree12-05-06 Lecture 27: Research Directions 21Trends: Growing Application Diversity Relay Node Access PointSensorWired InternetAd-Hoc/Sensor NetworksCollision Avoidance:Car NetworksWireless Home MultimediaMesh Networks12-05-06 Lecture 27: Research Directions 22Trends: Spectrum ScarcityPortland 8683 54San Diego 7934 76San Fran 3037 85Boston 2551 39#APs Max @ 1 spot•Densities of unlicensed devices already high•Spectrum is scarce will get worse•Improve spectrum utilization (currently 10%)12-05-06 Lecture 27: Research Directions 23Implications: Spectrum Scarcity•Interference and unpredictable behavior•Need better management/diagnosis tools•Lack of isolation between deployments•Cross-domain and cross-technologyWhy is my 802.11 not working?12-05-06 Lecture 27: Research
View Full Document