Network Computing Laboratory 1 Vivaldi: A Decentralized Network Coordinate System Authors: Frank Dabek, Russ Cox, Frans Kaashoek, Robert Morris MIT Published at SIGCOMM ‘04

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 2 Vivaldi is a simple, adaptive, distributed algorithm for co mputing network coordinates that accurately predict Inter net latencies without requiring any fixed infrastructure Internet Hosts compute their coordinates in some coordi nate space such that the distance between themselves a nd other host’s coordinates predicts the RTT between th em without sending probing packets just for measuring latencies Introduction

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 3 Algorithm Use synthetic distance between nodes to accurately map to latencies (RTT) between nodes. Can not create an exact mapping due to violations of triangle inequality N1 N2 N3 A B C || A || <= || B || + || C||

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 4 Algorithm Use synthetic distance between nodes to accurately map to latencies (RTT) between nodes. Can not create an exact mapping due to violations of triangle inequality N1 N2 N3 100 ms 48 ms 100 <= <= 96

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 5 Algorithm Use synthetic distance between nodes to accurately map to latencies (RTT) between nodes. Can not create an exact mapping due to violations of triangle inequality Tries to minimize the error of predicted RTT values Observation Minimizing the square error function of predicted RTT between two nodes is analogous to minimizing the energy in a mass-spring system Where: L ij = Actual Measure RTT between Node j and Node j x i = Synthetic coordinates of Node i x j = Synthetic coordinates of Node j   ij jiij x ||xLE 2 )||(

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 6 Algorithm N1N2 100

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 7 Algorithm N1N2 150 F 12 = 100 – 150 F 12 = -50

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 8 Algorithm F 12 = 100 – 100 F 12 = 0 N1N2 100

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 9 Algorithm N1N2 100 N3 100

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 10 Algorithm N3N2 150 F 32 = 100 – 150 F 32 = 50 N2 100 F 12 = 0 System Error = F F 32 2 = = 2500

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 11 Algorithm N1N2 150 F 12 = 100 – 150 F 12 = -50 N3 100 F 32 = 0 System Error = F F 32 2 = = 2500

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 12 Algorithm N3N1 125 F 32 = 100 – 125 F 32 = 25 N2 125 F 12 = 100 – 125 F 12 = 25 System Error = F F 32 2 = = 1300

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 13 Centralized version Calculate net Force on node i Move a step in the direction of the net Force

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 14 δ setting Sampling inaccurate node will not significantly affect the movement

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 15 Full Vivaldi Version Adjust time step

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 16 Impact of Adaptive time stamp

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 17 Evaluation Methodology Environment Packet-level network simulator using measured RTT values from the Internet Latency data Matrix of inter-host Internet RTTs Compute coordinates from a subset of these RTTs Check accuracy of algorithm by comparing simulated results to full RTT matrix 2 Data sets (Measured Data) 192 nodes Planet Lab network, all pair-ping gives fully populated matrix Median RTT = 76 ms 1740 Internet DNS servers Median RTT = 159 ms populate full matrix using the King method Continuously measure pairs over a week take median (other schemes just keep minim measured RTT since King can give estimates that are lower than actual RTT need to take median) During collection of data need to make sure unwanted forwarding of name request did not occur (give RTT for the wrong name server)

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 18 Evaluation Methodology 2 Data sets (Synthetically generated Data) Grid Vivaldi accurately recovers RTT values but coordinates are translated and rotated from the original grids coordinates ITM topology generation

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 19 Using data Simulation test setup Input RTT matrix Send a packet one a second Delay by ½ RTT time Send RPC packet Uses measured RTT of RPC to update coordinates

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 20 Evaluation (Robustness to high error nodes) Adding many new nodes that do not know their coordinates s, so are very uncertain (200 stable, then 200 new) Constant delta, already certain node get knock away from there good coordinates Adaptive delta, already certain nodes stay stable while new nodes move relatively quickly to their correct coordinates

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 21 Evaluation (Communication Patterns) In 21 (localization in sensor networks) shown that sampling only low latency nodes gives good local coordinates but poor global coordinates. 400 node sim (set 4 close neighbor, set 4 far neighbor) chose from far neighbor set is a probability p. p =.5 quick convergence p >.5 convergence slows p <.5 convergence slows no distant communication

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 22 Evaluation (Adapt to network changes) Ability to adapt to changes in the network (tested with “Transit-Stub”) extend one stub by 10x

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 23 Evaluation (Accuracy vs. GNP)

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 24 Model Selection

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 25 Critique Cool points Proposes simple & decentralized way for measuring network latency Fast convergence Bad points Limited scope of application area due to its dependency on traffic pattern Applications communicating neighbors are less benefited from Vivaldi The implication of delta(δ) is profound but no guidance provided No proposed architecture for managing coordinates information

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 26 New idea Scalable and distributed VC management system SPoN control plane? Exploits vivaldi scheme in clustering operations Cluster member nodes have received VC of other head nodes They decide to migrate the cluster by computing the distance between them and the cluster head nodes Exploits the spring-relaxation scheme in selecting data delivery node and clustering operatios in SPoN cluster Construct n dimensional virtual space with data loss, network bandwidth and latency. Compute the best VC in terms of the minimization Select the real node deployed in that coordinates or close to it

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 27 Backup slides

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 28 Evaluation Methodology

Korea Advanced Institute of Science and Technology Network Computing Laboratory | 29 Using data Error definitions Error of Link Absolute difference between predicted RTT (coordinate math) and measured (RTT Matrix element) Error of Node Median of link errors involving this node Error of System Median of all node errors