A New Path Selection Algorithm for MPS Networks Based on Available Bandwidth Measurement T. Anjali, C. Scoglio, J. de Oliveira,. C. Chen, I. F. Akyildiz, J. A. Smith *, G. Uhl *, A. Sciuto * Broadband and Wireless Networking aboratory School of Electrical and Computer Engineering Georgia Institute of Technology * ESDIS Network Prototyping aboratory NASA Goddard Space Flight Center 1
Summary Need for Measurement Available Bandwidth Measurement Path Selection Algorithm Available Bandwidth Estimator Experimental Results 2
Need for Measurement QoS-aware path selection for Efficient Traffic Engineering User QoS satisfaction Efficient path selection finds a path with sufficient resources Requires knowledge of resource availability in the network 3
Capacity vs. Available Bandwidth Capacity: Maximum throughput without crosstraffic Available bandwidth: Maximum throughput given cross-traffic A 1 A 2 A 3 A 4 Source Destination C 1 C 2 C 3 C 4 4
Existing Methods Pathchar based (Jacobson) Packet pair based (Bolot) Nettimer (ai and Baker) AMP (NANR) OCXmon (NANR) MRTG (Oetiker) Pathload (Dovrolis and Jain) 5
MRTG MRTG to monitor the traffic load on network links Highly portable SNMP based tool Provides only 5 min averages of link utilization Used by the network operators only as router SNMP community string information required ink-by-link measurement 6
Pathload Sends Self-oading Periodic Streams at increasing rates till the rate is larger than the tight link available bandwidth and the relative one way delays of packets show an increasing trend. This scheme is highly intrusive even though the scheme measures the available bandwidth of the tight link End-to-end available bandwidth measurement 7
Path Selection Algorithm Various algorithms for path selection Guerin/Orda (GOBECOM 1997) Matta/Bestavros (INFOCOM 1998) Kar/Kodialam (JSAC 2000) Require nominal available bandwidth information Actual traffic does not follow the SA agreement Non-efficient network resource utilization 8
Path Selection Algorithm (Contd.) More accurate link utilization information Not very frequently Scalabilty Correctness ink utilization estimation and prediction 9
Path Selection Algorithm (Contd.) 1. At time instant k, a bandwidth request r arrives between nodes i and j. 2. Run the available bandwidth estimation algorithm on links with no bandwidth estimation available. 3. Compute the best path using the shortest widest path algorithm with weights as computed in Step 2. 4. Obtain the available bandwidth A on the bottleneck link of the path. 5. If r>a*threshold, reject this path and return to step 3. Else, the path is selected for the request. 6. If no path is available, request is rejected and network is congested. 10
Available Bandwidth Estimation Method Accurate, reliable estimate of the available bandwidth of a link Based on MRTG++ Reliably predicts the utilization of the link for a future interval, that varies in size 11
Available Bandwidth Estimation (Contd.) k-p+1 k k+h We use the past p samples to predict the utilization for the next h samples Utilize the Wiener-Hopf equations for prediction Values of p and h varied according to the estimation error 12
13 Available Bandwidth Estimation (Contd.) inear prediction problem: Wiener-Hopf equations (solved by evinson recursion) = = + 1 0 ] [1, for ] [ ] [ ] [ p n a h a n w n k a k τ τ + = 1) ( ) ( 1) ( (0) (0) 1) ( 1) ( (0) p a r a r p w w r p r p r r a a M M M O M
Available Bandwidth Estimation (Contd.) 1. At time instant k, available bandwidth measurement is desired 2. Find the prediction coefficients in the Wiener - Hopf equations using evinson recursion. 3. Find predicted utilization from the linear prediction 4. Predict available bandwidth from the utilization 5. Find prediction error 6. Obtain new values for p and h based on the prediction error. 14
Performance Evaluation Available Bandwidth Estimator 15
Performance Evaluation (Contd.) Available Bandwidth Estimator 16
Performance Evaluation (Contd.) Simulation Network Topology 1 2 23 3 12 44 77 64 4 36 5 11 SP request from node 2 to 8 Traditional shortest path scheme selects path 2 fi 12 fi 11 fi10 fi9 fi8 Proposed scheme selects path 2 fi3 fi4 fi5 fi6 fi7 fi8 which has less nominal available bandwidth but larger actual available 43 bandwidth, which is measured by our available bandwidth estimato r 74 10 68 6 9 30 40 7 57 8 17
Performance Evaluation (Contd.) Rejection probability for node pair 2-8 with increasing traffic load on the node pair, in presence of background traffic. 18
Conclusions and Future Work Proposed a new QoS-aware path selection algorithm Utilizes a scheme for accurate estimations of link available bandwidth Based on MRTG++ Utilizes Wiener-Hopf equations for prediction Tells the duration for which the prediction is valid Achieved lower rejection probability due to more accurate network state information Future Work: Multi-domain path selection scheme 19