Large Scale WAN Emulation
Total Page：16
File Type：pdf, Size：1020Kb
Large Scale WAN Emulation Martin Arlitt Rob Simmonds Carey Williamson - University of Calgary Calgary Alberta March 18, 2002 Outline • Overview of WAN Simulation & Emulation • Introduction to IP-TNE • Discussion of Related Projects • Validation of IP-TNE • Current Projects Involving IP-TNE • Future Work 1 Performance Evaluation Approaches 1. Experimental + offers the most realistic environment - requires significant financial investment - can be difficult to repeat results - restricted to existing technologies 2 Performance Evaluation Approaches 2. Simulation + low-cost, flexible, controllable, reproducible environment - abstractions can compromise usefulness of results 3. Analytical + provides quick answers - often requires the greatest degrees of abstraction 3 Performance Evaluation Approaches 4. Emulation * a hybrid performance evaluation methodology * combines aspects of other three approaches + enables controlled experimentation with existing applications - still suffers from drawbacks of other approaches 4 Wide-Area Network Simulation • provides a virtual Wide-Area Network (WAN) environment • allows all network conditions to be controlled – packet loss – packet reordering/duplication – link bandwidths – propogation delays – asymetric links – bounded queue sizes – multipath • allows alternative networking technologies to be evaluated 5 Wide-Area Network Emulation • extends capabilities of WAN simulation • enables controlled testing with unmodified applications • both simulation and emulation are important tools 6 Challenges • scaling to large, non-trivial sized networks • developing realistic models of network characteristics • keeping pace with real world events (emulator) 7 Introduction to IP-TNE IP-TNE: Internet Protocol Traffic and Network Emulator • developed at the University of Calgary • a collaborative effort between researchers in: – high performance parallel simulation (Rob) – IP networking (Carey) 8 Introduction to IP-TNE (cont.) Design • built around IP-TN network simulator • uses a high performance PDES kernel • kernel modified to provide real-time features • can simulate end hosts as well 9 Introduction to IP-TNE (cont.) Operation • reads packets promiscuously • replies to ARP requests for registered hosts • packets timestamped when read from network • output time calculated by network simulator • packets released when wall-clock time = output time 10 Introduction to IP-TNE (cont.) Current Strengths • high performance – this enables detailed simulation of non-trivial networks • flexibility Current Weaknesses • simple models of network behaviour • visualizing operation of large networks • need for repeatable validation tests 11 Related Projects ns-2 • most well known network simulator/emulator + feature rich (e.g., multiple flavours of TCP, etc.) - poor performance restricts size of simulated network SSFNet + focused on simulating extremely large scale networks - does not provide emulation capability 12 Related Projects (cont.) Other Network Simulators/Emulators • dummynet • NistNet • Utah Network Testbed • Network Emulator 4.0 • The Ohio Network Emulator (ONE) • Opnet • Shunra Cloud 13 Validation of IP-TNE Overview • validation is not a one-time event, but a continuous process Protocol Implementation • capture and examine network traffic e.g., use tcpdump to capture packets between emulator and host – examine HTTP headers and responses for correctness – examine IP and TCP headers for correctness – analyze TCP behaviour for correctness 14 Validation of IP-TNE (cont.) Emulator Behaviour • compare application behaviour in real/emulated environments e.g., Web server benchmarking – evaluate Web server performance with benchmarking tools – compare to benchmarking results using IP-TNE – account for any differences; repeat tests if necessary – can also compare to results from other emulators 15 Current Projects Web Server Benchmarking • demonstrate one use of IP-TNE • helps validate our emulator (comparing to Nahum et al.) • show that a centralized approach to WAN emulation: – can scale to large network sizes – is more manageable than a distributed approach 16 Current Projects (cont.) Alberta SuperNet • what is SuperNet? – ICT infrastructure throughout province of Alberta – will provide high-speed Internet access to: ∗ 100% schools ∗ 95% businesses ∗ 80% residences – 422 communities to be linked – 5,175 miles of fiber; wireless networks in remote areas 17 Current Projects (cont.) Alberta SuperNet • IP-TNE will be used in development of SuperNet • data captured from SuperNet will be used to: – develop more realistic models of network conditions for IP-TNE – validate IP-TNE • project will test scalability of IP-TNE 18 19 Future Work • improve models of network conditions • add additional features (e.g., TCP variants, routing algs) • make protocol validation a repeatable/automated process • improve real-time performance of PDES kernel • participation in SuperNet • use ELISA (an Internet testbed) to further improve IP-TNE 20 Contact Information Martin Arlitt: [email protected] Rob Simmonds: [email protected] Carey Williamson: [email protected] IP-TNE: http://warp.cpsc.ucalgary.ca/ 21 SuperNet Community Rollout Schedule - 2002 Bruderheim Lamont Fort Saskatchewan Chipman Mundare Enoch Nisku Devon Innisfree Lindale Beaumont Calmar Mannville Drayton Valley Thorsby Breton Warburg Millet Alder Flats Winfield Buck Lake Bentley College Heights Leslieville Clive Eckville Rocky Mountain House Blackfalds Sylvan Lake Condor Benalto Scheduled Community completion (Calendar Year) 2002 2003 2004 NOTE: Not all communities are labelled. 4 February, 2002. SuperNet Community Rollout Schedule Zama City Meander River Fort Chipewyan Chateh Garden Creek Rainbow Lake High Level Fox Lake Fort Vermilion La Crete Buffalo Head Prairie Paddle Prairie Tall Cree I.R. 173 Keg River Fort MacKay Manning Chipewyan Lake Fort McMurray Peerless Lake Loon Lake Worsley Dixonville Trout Lake Cadotte Lake Anzac Cleardale Bear Canyon Hines Creek Peace River Berwyn Silver Valley Fairview Nampa Bonanza Atikameg Wabasca-Desmarais Chard Eaglesham Sandy Lake Rycroft Donnelly Conklin Woking Grouard Valhalla Horse Lake I.R. 152BTeepee Creek Driftpile Slave Lake Beaverlodge Bezanson East Prairie Metis Settlement Smith Wandering River Elmworth Grovedale Sturgeon Lake I.R. 154 Heart Lake Lac La Biche Swan Hills Flatbush Athabasca Boyle Fawcett Buffalo Lake Iron River Cold Lake Fox Creek Rochester Dapp Ardmore Glendon Blue Ridge Barrhead Clyde Thorhild Smoky Lake Ashmont Elizabeth Metis Settlement Mayerthorpe Elk Point Grande Cache Alexander I.R. 134Bruderheim Andrew Lafond Lamont Heinsburg Fort Saskatchewan Chipman Peers St. Albert Myrnam Entwistle Edmonton Mundare Edson Ardrossan Clandonald Vegreville Enoch Nisku Hinton Devon Tofield Innisfree Islay Lindale Beaumont Calmar Lloydminster Drayton Valley Mannville Thorsby Hay Lakes Bruce Breton Warburg Millet Paradise Valley Wetaskiwin Alder Flats Falun Bawlf Irma Jasper Winfield Wainwright Rosalind Killam Buck Lake Bluffton Mecca Glen Hardisty Chauvin Donalda Forestburg Bentley College Heights Amisk Leslieville Clive Eckville Alliance Czar Rocky Mountain House Blackfalds Botha Brownfield Cadogan Sylvan Lake Condor Castor Benalto Penhold Bodo Coronation Big Valley Elnora Consort Bowden Altario Rumsey Sundre Olds New Brigden Morrin Delia Hanna Carstairs Youngstown AcmeCarbonDrumheller Beiseker Cereal Airdrie Rockyford Banff Bearspaw Acadia Valley Scheduled Community completion Exshaw Strathmore Hussar Cessford Bragg Creek Gem (Calender Year) Carseland Bindloss Bassano Arrowwood Duchess Jenner Black Diamond Milo Brooks Longview Brant Tilley 2002 Vulcan Nanton Lomond Rainier Schuler Champion Ralston Stavely Carmangay 2003 Hays Redcliff Barons Dunmore Granum Picture Butte Bow Island Coalhurst Barnwell 2004 Bellevue Brocket Wrentham Blood I.R. 148 Foremost Manyberries Glenwood Warner Cardston Milk River NOTE: Not all communities are labelled. 4 February, 2002. Coutts.