Estimating Link Capacity in High Speed Networks LingJyh

  • Slides: 20
Download presentation
Estimating Link Capacity in High Speed Networks Ling-Jyh Chen 1, Tony Sun 2, Li

Estimating Link Capacity in High Speed Networks Ling-Jyh Chen 1, Tony Sun 2, Li Lao 2, Guang Yang 2, M. Y. Sanadidi 2, Mario Gerla 2 1 Institute of Information Science, Academia Sinica 2 Dept. of Computer Science, University of California at Los Angeles

Definition n n Capacity: Capacity maximum IP-layer throughput that a flow can get, without

Definition n n Capacity: Capacity maximum IP-layer throughput that a flow can get, without any cross traffic. Available Bandwidth: Bandwidth maximum IP-layer throughput that a flow can get, given (stationary) cross traffic.

Previous Work on Capacity Estimation n Per-hop based ¨ pathchar: use different packet sizes

Previous Work on Capacity Estimation n Per-hop based ¨ pathchar: use different packet sizes to probe the perhop link capacity ¨ clink, pchar: variants of pathchar ¨ Nettimer: use “packet tailgating” technique n End-to-end based ¨ Pathrate, n n Sprobe, Cap. Probe For specialized networks: Asym. Probe, ALBP, Ad. Hoc Probe How about high speed networks?

Estimating High Speed Links n High speed links are becoming popular (e. g. GB

Estimating High Speed Links n High speed links are becoming popular (e. g. GB links, DVB links, and UWB links) n However, capacity estimation on high speed links remains a challenge (e. g. , probing pksize and system time resolution are limited) n And, an effective estimation tool for high speed links is still desired

Our Contribution n We propose an end-to-end capacity estimation technique for high speed links,

Our Contribution n We propose an end-to-end capacity estimation technique for high speed links, called PBProbe n PBProbe is based on Cap. Probe ¨ One-way method ¨ UDP based ¨ packet bulk based ¨ simple, fast, and accurate

Packet Pair Dispersion T 1 T 3 Narrowest Link T 2 20 Mbps 10

Packet Pair Dispersion T 1 T 3 Narrowest Link T 2 20 Mbps 10 Mbps T 3 5 Mbps. T 3 10 Mbps T 3 20 Mbps Capacity = (Packet Size) / (Dispersion) 8 Mbps

Issues: Compression and Expansion • Queueing delay on the first packet => compression •

Issues: Compression and Expansion • Queueing delay on the first packet => compression • Queueing delay on the second packet => expansion

Cap. Probe (Rohit et al, SIGCOMM’ 04) n n Key insight: a packet pair

Cap. Probe (Rohit et al, SIGCOMM’ 04) n n Key insight: a packet pair that gets through with zero queueing delay yields the exact estimate. Cap. Probe uses “Minimum Delay Sum” filter. Capacity

Proposed Approach: PBProbe Have two phases for both forward and backward link estimation n

Proposed Approach: PBProbe Have two phases for both forward and backward link estimation n Use packet bulk (instead of packet pair) of length k in each probing n Adapt k to enlarge the dispersion between the first and last packet, and thus overcome the timer resolution problem n Tradeoff BW consumption and estimation speed by U parameter n

Proposed Approach: PBProbe

Proposed Approach: PBProbe

Proposed Approach: PBProbe n k is depended on the estimated link capacity and the

Proposed Approach: PBProbe n k is depended on the estimated link capacity and the supported timer resolution. n n is set to 200. Dthresh is set to 1 ms. U is set to 0. 002. n n

Analysis Poisson cross traffic (arrival and service rates are λ and μ), service time

Analysis Poisson cross traffic (arrival and service rates are λ and μ), service time is τ n Prob. of obtaining a good sample: n n Expected number of samples required for obtaining a good sample:

Analysis

Analysis

Evaluation NISTNet emulation n High speed Internet experiments n Comparison of PBProbe and Pathrate

Evaluation NISTNet emulation n High speed Internet experiments n Comparison of PBProbe and Pathrate n

Evaluation 1: NISTNet emulation n No cross traffic

Evaluation 1: NISTNet emulation n No cross traffic

Evaluation 2: Internet experiments n 5 hosts: NTNU, UCLA, Cal. Tech, Ga. Tech, PSC

Evaluation 2: Internet experiments n 5 hosts: NTNU, UCLA, Cal. Tech, Ga. Tech, PSC (n = 200, k = 100, 20 runs)

Evaluation 3: PBProbe vs Pathrate

Evaluation 3: PBProbe vs Pathrate

Summary We propose an end-to-end capacity estimation technique, called PBProbe, for high speed links.

Summary We propose an end-to-end capacity estimation technique, called PBProbe, for high speed links. n We evaluate PBProbe by analysis, emulation and Internet experiments. n We show that PBProbe can correctly and rapidly estimate bottleneck capacity in almost all test cases. n

Acknowledgements n This work is co-sponsored by the National Science Council and the National

Acknowledgements n This work is co-sponsored by the National Science Council and the National Science Foundation under grant numbers NSC-94 -2218 E-001 -002 and CNS-0435515. n We are grateful to the following people for their help in carrying out PBProbe measurements: Sanjay Hegde (Cal. Tech), Che-Chih Liu (NTNU), Cesar A. C. Marcondes (UCLA), and Anders Persson (UCLA).

Thanks! Cap. Probe: http: //nrl. cs. ucla. edu/Cap. Probe/

Thanks! Cap. Probe: http: //nrl. cs. ucla. edu/Cap. Probe/