Overlay NetworksHow Robust is Internet Routing?RON: Routing Using OverlaysRON DesignRON greatly improves loss-rateAn order-of-magnitude fewer failuresResilience Against DoS AttacksThroughput ImprovementOverlay NetworksHari Balakrishnan6.829 Computer NetworksHow Robust is Internet Routing?1. Slow outage detection and recovery2. Inability to detect badly performing paths3. Inability to efficiently leverage redundant paths4. Inability to perform application-specific routing5. Inability to express sophisticated routing policyPaxson 95-97• 3.3% of all routes had serious problemsLabovitz 97-00 • 10% of routes available < 95% of the time• 65% of routes available < 99.9% of the time• 3-min minimum detection+recovery time; often 15 mins• 40% of outages took 30+ mins to repairChandra 01 • 5% of faults last more than 2.75 hoursRON: Routing Using Overlays•Cooperating end-systems in different routing domains can conspire to do better than scalable wide-area protocols•Types of failures–Outages: Configuration/op errors, software errors, backhoes, etc.–Performance failures: Severe congestion, DoS attacks, etc.Scalable BGP-based IP routing substrateReliability via path monitoringand re-routingReliability via path monitoringand re-routingReliability via path monitoringand re-routingReliability via path monitoringand re-routingRON DesignProberProberRouterRouterForwarderForwarderConduitConduit Link-state routing protocol, disseminates info using RON!PerformanceDatabaseApplication-specific routing tablesPolicy routing moduleRON libraryNodes in differentrouting domains (ASes)00.20.40.60.810 0.2 0.4 0.6 0.8 1"loss.jit"RON greatly improves loss-rate30-min average loss rate with RON30-min average loss rate on Internet13,000 samplesRON loss rate never more than 30%An order-of-magnitude fewer failuresLoss RateRON BetterNo ChangeRON Worse10% 479 57 4720% 127 4 1530% 32 0050% 20 0080% 14 00100% 10 0030-minute average loss rates6,825 “path hours” represented here12 “path hours” of essentially complete outage76 “path hours” of TCP outageRON routed around all of these!One indirection hop provides almost all the benefit!6,825 “path hours” represented here12 “path hours” of essentially complete outage76 “path hours” of TCP outageRON routed around all of these!One indirection hop provides almost all the benefit!Resilience Against DoS AttacksThroughput Improvement00.10.20.30.40.50.60.70.80.910.1 1 10Fraction of samplesRatio of RON throughput to direct throughput (logscale)2x increase from RON(109 samples)1/2x decrease from RON(20 samples)bw