FlowSense: Monitoring Network Utilization with Zero Measurement Cost Curtis Yu1, Cristian Lumezanu2, Yueping Zhang2, Vishal Singh2, Guofei Jiang2, Harsha V. Madhyastha1 1UC Riverside, 2NEC Labs America SDN enables centralized control 2 SDN enables reactive control of network B A C 3 State-of-the-art measurement Active measurements Injection of probes Passive measurements Expensive instrumentation and infrastructure setup SDN measurements Additional control traffic 4 FlowSense: measuring with zero cost Leverage existing control traffic to measure network No additional traffic introduced Network informs system of changes As accurate as switch polling 5 OpenFlow overview Flow Arrival: 1) Packet of new flow arrives 2) PacketIn message sent 3) FlowMod message sent 4) New rule installed sw1 Flow Completion: 1) Rule expires 2) Remove rule 3) FlowRemoved message sent Rule Match Output 1 Green sw2 sw2 6 Messages have utilization info PacketIn Signals new flow on port (t1) FlowRemoved Duration of entry in flow table (10s) Amount of traffic matched (400 MB) Can infer utilization contributed by flow on link t1 t1 + 10s PacketIn FlowRemoved Flow: 7 Estimating utilization in FlowSense Time t1 t2 t3 t4 t5 t6 f1 f2 f3 Checkpoint t4 Flows Left 2 1 0 Total Utilization 40 MB/s 50 70 t5 t6 0 1 0 30 MB/s 10 20 MB/s 8 FlowSense system design Parser Captures and analyzes control traffic Utilization Monitor Runs FlowSense algorithm Utilization Table Database of checkpoints 9 Evaluating FlowSense Goals Accuracy Timeliness Dataset Testbed of NEC OpenFlow switches Real-world traffic trace from a campus network switch 10 Accuracy testbed 1 continuous flow 3 separate short flows 11 FlowSense is as accurate as polling Utilization (MBps) 60 Polling FlowSense 50 40 30 20 10 0 0 50 100 150 Time Elapsed (sec) 200 12 Revisiting utilization estimation Time t1 t2 t3 t4 t5 t6 f1 f2 f3 Checkpoint t4 Flows Left 0 2 Total Utilization 70 MB/s 40 t5 t6 0 0 30 MB/s 20 MB/s 13 Perfect utilization information is typically delayed Fraction of checkpoints 1 0.8 0.6 0.4 0.2 1 10 100 1000 Time until last active flow ends (s) 14 Coarse-grained estimation can be obtained soon 15 Data refreshes in the order of seconds 16 Conclusion FlowSense Leverage control traffic to estimate utilization Incurs no measurement cost Estimates are accurate Comparable to switch polling Data is timely Do not have to wait too long for new utilization information 17 Future work Measuring long flows Overhead of stopping the rule for measurement Wildcard and proactive rules Reasonable way to merge active measurements to keep overhead low Link Latency Similar method to enable zero cost measurement of latency 18 Questions? 19