Bandwidth Estimation Workshop 2003

Evaluating pathrate and pathload with realistic cross-

traffic

Ravi PrasadManish Jain

Constantinos Dovrolis (ravi, jain, dovrolis@cc.gatech.edu)

College of ComputingGeorgia Institute of Technology

Bandwidth Estimation Workshop 2003

Background

PathrateEstimates path capacityBased on packet pair/train dispersion

Packet pair estimates: Set of possible capacity modes

Packet train estimates: ADR=Lower bound on capacity

Capacity = (Strongest and narrowest mode > ADR)

PathloadEstimates path available bandwidth (avail-bw)Based on one-way delay trend of periodic

streamsReports a range of avail-bw

Corresponds to variation in stream durationhttp://www.pathrate.org

Bandwidth Estimation Workshop 2003

Motivation

Recent studies pointed towards poor accuracy of these toolshttp://www.caida.org/outreach/presentations/2

003/bwest0308/doereview.pdfA measurement study of available bandwidth

estimation tools. Strauss et. al. IMC 2003Our objective: re-evaluate accuracy of

both toolsWide range of cross-traffic loadRealistic cross-traffic Completely monitored testbed (no guessing!)

Bandwidth Estimation Workshop 2003

Outline

Describe test methodologyTestbedCross-traffic type

Show accuracy results100Mbps path1Gbps pathWith Iperf cross-traffic

Explaining inaccuracies with Iperf cross-traffic

Conclusions

Bandwidth Estimation Workshop 2003

Testing methodology

Used local testbedComplete knowledge of path properties

Capacity Available bandwidth

Complete control of cross-traffic Rate Type (TCP vs UDP vs trace-driven)

Bandwidth Estimation Workshop 2003

Testbed

Narrow link capacity C = 100Mbps or 1Gbps

Bandwidth Estimation Workshop 2003

Cross traffic

Trace-driven cross-traffic generation: NLANR traces

OC-3, OC-12, OC-48 Trace information at the end of the talk

Packet size distributionUnmodified

Packet interarrivalsEither, scaled to achieve desired cross-traffic throughputOr, unmodified

Iperf-based cross-traffic Single TCP stream UDP stream

Bandwidth Estimation Workshop 2003

Results

Bandwidth Estimation Workshop 2003

FastEthernet: Traces with scaled interarrivals

Bandwidth Estimation Workshop 2003

FastEthernet:Traces with unmodified interarrivals

Bandwidth Estimation Workshop 2003

Gigabit path:Traces with scaled interarrivals

Bandwidth Estimation Workshop 2003

Gigabit Path:Iperf UDP Cross traffic

Bandwidth Estimation Workshop 2003

Gigabit path:Iperf single stream TCP

Bandwidth Estimation Workshop 2003

Unrealistic cross-traffic

Single stream TCP Entire window appears as

burst at beginning of RTT Minimum averaging

interval: RTT

L/C

TR

Tw

UDP periodic stream

Packet size: L Rate: R Dispersion: L/R Utilization = R/C

Bandwidth Estimation Workshop 2003

TR

TW TO

Seeks some “off” time periods of duration larger than L/C L: Probe size

TCP traffic Off period TO =TR - TW- L/C

Correct capacity estimate when TO > L/C

Pathrate under unrealistic traffic

L/C UDP periodic traffic If < 0.5then TO > L/C Else, underestimation

Bandwidth Estimation Workshop 2003

TRTW

Pathload under unrealistic traffic

Samples avail-bw in stream duration (TS)

TCP traffic Avail-bw averaging period TR

If TS << TR then:

Wide Avail-bw range estimate

UDP periodic traffic Avail-bw averaging period

L/R TS = 100 x L/C > L/R Correct avail-bw range

estimate

TSTS TS

L/C TS

Bandwidth Estimation Workshop 2003

Conclusions

Type of cross-traffic is important for bandwidth estimation tools

Pathrate and pathload perform well with realistic cross-traffic

Simulated traffic does not capture:Packet size distributionInterarrival distributionCorrelation structure

Bandwidth Estimation Workshop 2003

Trace identifiers

OC3 : MEM-1070464136-1, COS-1070488076-1, BWY-1063315231-1,

COS-1049166362-1OC12: MRA-1060885637-1OC48: IPLS-CLEV-20020814-093000-1