1. Sponsored by the National Science Foundation Systematic Experimentation Case Study: Virtual Router Failure Restoration Xuan Liu UMKC/GENI Project Office

2. Sponsored by the National Science Foundation 2 Systematic Experimentation – GEC 20 – June 23, 2014 Motivation Design Failure Restoration Scheme Design Optimization Model Design Optimization Model Heuristic Algorithm Deploy Experiments On GENI Deploy Experiments On GENI Goal: Dynamically recover multiple virtual router failures using standby virtual routers in a virtual network environment. Collect Data (bw, delay, etc) Collect Data (bw, delay, etc) Virtual network Find the optimal Standby Virtual Router Complex Topology requires systematic experimentation setup

3. Sponsored by the National Science Foundation 3 Systematic Experimentation – GEC 20 – June 23, 2014 Prior Work Experimental Study on GpENI Testbed –Test the conceptual dynamic reconfiguration scheme for virtual router failure restoration –Lessons learned The proposed scheme is practical in a distributed network Substrate network topology has impact to the delay between virtual nodes GpENI Network: US SitesGpENI Network: European Sites

4. Sponsored by the National Science Foundation 4 Systematic Experimentation – GEC 20 – June 23, 2014 Why Choosing GENI Testbed A distributed virtual network testbed is desired to deploy the heuristic multi- criteria S-VR selection algorithm, so that the impact to the virtual networks can be better understood –GENI Testbed provides such capabilities. Geographically distributed Customized software installation on VMs is allowed Arbitrary virtual topology creation Backbone topology is more general GpENI NetworkGENI Network

5. Sponsored by the National Science Foundation 5 Systematic Experimentation – GEC 20 – June 23, 2014 Target Topology on GENI Stitched Links between aggregates LANs within single aggregates End Hosts Virtual Routers (Colors shows different geographical locations)

6. Sponsored by the National Science Foundation 6 Systematic Experimentation – GEC 20 – June 23, 2014 Systematic Experimentation Design Install Software on One Node Automate Experimental Configuration End-to-End Validation Scaling Up Automate Measurement Use Smallest Topology Increases nodes/sites Important Tips: Make One Change at A Time Version Control

7. Sponsored by the National Science Foundation 7 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #1 Install Software on One Node MANUAL Install XORP (took > 1 hour) Download XORP software Install all required dependences $apt-get install libssl-dev $apt-get install ncurses-dev $apt-get install libpcap-dev $apt-get install g++ $apt-get install traceroute $apt-get install scons Compile XORP Build XORP $sudo groupadd xorp AUTOMATE Create custom image VMrouter Load the URL of the Custom image How to Create Custom Image: http://groups.geni.net/geni/wiki/HowTo/ManageCustomImagesInstaGENI Install XORP on one Node Create Custom image Save the URL to the Custom Image

8. Sponsored by the National Science Foundation 8 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #2 Automate Configuration MANUAL Configure OSPF and starts XORP AUTOMATE: Configures OSPF and starts XORP route r Starts with smallest reasonable topology Get Familiar with the Syntax of the Configuration File Validate the OSPF Routing in the Topology Validate the OSPF Routing in the Topology Write Shell Script to Automate OSPF Configuration Procedure Load the Automation Script as a Postboot Script in Flack to each VM One pair of nodes are not adjacent Each node has at least two interfaces router

9. Sponsored by the National Science Foundation 9 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #3 End-to-end Validation route r client server route r Manually Install Iperf to client/server Validate End-to-End Communication Validate End-to-End Communication Add Client and Server Take similar procedure in step #2 to automate iperf installation using a postboot script.

10. Sponsored by the National Science Foundation 10 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #4 Measurement Track routing table updates – MANUAL Script to track routing table updates – AUTOMATE Periodically Capture Routing table Tracking routing table updates by killing XORP process (mimic virtual router failure) – Tracking the routing table changes using terminal – Display graphs of routing table changes using Labwiki Followed step #1 to create a new custom image with both LabWiki and XORP installed. route r client server

11. Sponsored by the National Science Foundation 11 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #3 End-to-End Validation router-1 route r router-2 router-3 client server Outputs from the server:

12. Sponsored by the National Science Foundation 12 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #5 Scaling up Scale to more aggregates TOPOLOGY Stitch between 2/3 aggregates TEST Repeat earlier tests route r client server route r client server

13. Sponsored by the National Science Foundation 13 Systematic Experimentation – GEC 20 – June 23, 2014 Single Aggregate: Multiple Aggregates with Stitched Links: Case Study #5 Scaling up

14. Sponsored by the National Science Foundation 14 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #5 Scaling Up Scale nodes at aggregate TOPOLOGY Add more routers and links TEST Repeat earlier tests route r client server route r client server

15. Sponsored by the National Science Foundation 15 Systematic Experimentation – GEC 20 – June 23, 2014 router-3router-1router-2 router-4 router-7router-6 router-5 client server Traceroute from server to client Iperf TCP test outputs from server

16. Sponsored by the National Science Foundation 16 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #5 Scaling Up Display the routing table update from three routers over time in Labwiki

17. Sponsored by the National Science Foundation 17 Systematic Experimentation – GEC 20 – June 23, 2014 Case Study #5 Scaling Up Scaling up at both dimensions (nodes and aggregates) route r client server route r client server route r client server To be done

18. Sponsored by the National Science Foundation 18 Systematic Experimentation – GEC 20 – June 23, 2014 Lessons Learned / Best Practices Lessons Learned –Learn tool first. Then use in experiment. –Underlying physical topology –Experience with GENI rack Challenges with debugging postboot script execution –Always start with a small postboot script execution Easy to get help from geni-user group mailing list Best Practices –Version control Software Scripts RSpec –Keep the RSpec (both request and manifest RSpec) at each step, so we can track the updates/errors

19. Sponsored by the National Science Foundation 19 Systematic Experimentation – GEC 20 – June 23, 2014 Demo