Presentation on theme: "Diverse Network Services and Remaining Challenges SRCCS Winter 2005 Workshop on Internet Modeling and Analysis Sue B. Moon Division of Computer Science."— Presentation transcript:
Diverse Network Services and Remaining Challenges SRCCS Winter 2005 Workshop on Internet Modeling and Analysis Sue B. Moon Division of Computer Science Dept. of EECS KAIST
2 Diverse Data Sets in Korea Data from major ISPs –Non-existent in public Data from academia –DAGMON traces at KAIST –Long-term, sampled flow-level traces at POSTECH –4-month-worth NetFlow traces at CNU Others –3G video streaming service data –2-day packet-level traces from a home network
3 Diverse Data Sets in Korea Data from major ISPs –Non-existent in public Data from academia –DAGMON traces at KAIST –Long-term, sampled flow-level traces at POSTECH –4-month-worth NetFlow traces at CNU Others –2-day packet-level traces from a home network –3G video streaming service data
4 2005.11.23 at KAIST FTP HTTP P2P Others P2P FTP In Bits/SecNo. of Pkts / sec
5 2004.12.9. at KAIST FTP HTTP P2P Others P2P In Bits/SecNo. of Pkts / sec
6 REMOVE after chat with SY Previous slides upstream or downstream? According to JH Youn –Upstream 70%/85% (Bytes/Pkts) p2p –Downstream Web and VOD more dominant that p2p.
7 Home Networking in Korea Architecture –Simple, tree-like topology –To each home = 100ME –Internal links = 1GE –Outbound speed = OC-3 or up Reality in new apt complexes –Control home appliances thru the net washing machine, gas stove, lights, heater/airconditioner, door lock by PDAs at home or remotely by cellphones/web access –Replace DSL/cable lines –Will be "backbone" for home ubiquitous sensor network Need for remote monitoring –Lack of resources
9 Video Streaming over 3G Goals of Monitoring –To satisfy every user: High revenue-generating customers More focus on per-user performance Challenges –E2E performance segmented over cellular and wired networks –No integrated monitoring solutions yet –No good metric for overall/single-user performance
10 Provisioning for Interactive Streaming Interactive Streaming –Not a driving force behind b/w –A candidate for growing revenue Examples –VoIP gradually taking over PSTN traffic –Remote video viewing at door by cell phone –Online game traffic "Good" routing more important than bandwidth
11 Routing in the Internet Intra-domain: shortest path routing Inter-domain: policy-based routing
12 Issues in "Good" Routing Misbehaving routing protocols –BGP misconfigurations –Pathological behaviors –Frequent changes Even under normal circumstances –Transient behaviors –Inter/intra-domain routing not well understood
13 Scenario for a Transient Routing Loop In Normal Operation
14 When a link fails, R1 is the first to detect.
18 VoIP experimental setup [Boutremans2002] Traffic injected in the network: –200 byte UDP packets –every 5ms. Packets captured and timestamped at end-systems. Traceroute runs continuously during the experiment.
19 Information Sources IS-IS & BGP listener logs Router logs from both ends of failing links Controlled bi-directional VoIP traffic between Reston and ATL SNMP data
20 Delays (1 sec timescale) 2 links down ~ 3.4ms 2 links up ~ 2.6ms 3 links down 3 links up
21 When the two interfaces went down … 6.6 seconds
22 When three links came back up … For 30 secs packets follow a shorter path Traffic black-holed for 1.745 seconds Traffic black-holed for 0.975 seconds
23 Approaches To Fix It Fine-tuning parameters –Timer values [Alattinoglu2002] Modify Routing Protocols –Suppress advertisement and perform local rerouting using a backwarding table [Lee04] –Centralized path computation [Feamster04,Rexford04]
24 Our Approach Key Idea: –Find disjoint overlay path and send duplicate packets Assumptions –Sender and receiver both within an AS –Bidirectional link weights –Extra income for extra b/w consumption Pros and cons –Advantages No modification to current infrastructure Selective use by only those that need it –Disadvantages Extra b/w consumption
29 Summary: Local Convergence Convergence delay during failures hurts more After IS-IS converges, extra delay due to FIB update Experimen t IS-IS Convergence Service Disruptions 2 links down5.1 sec6.6 sec 2 links up17.6 sec~ 0 sec 3 links down5.9 sec8.0 sec 3 links up17.5 sec0.95 sec
30 Causes and Open Questions Why only after 30 secs? –spf-interval set to 30secs What caused the 1.745 secs disruption? Unknowsn in GSR architecture What is the right timescale to propagate good news?