A Hidden Markov Model Approach to Available Bandwidth Estimation and Monitoring

1
A Hidden Markov Model Approach to
AvailableBandwidth Estimation and Monitoring
Cesar D. Guerrero Miguel A. Labrador Department
of Computer Science and Engineering University of
South Florida guerrerc, labrador_at_cse.usf.edu
2
Outline

• Motivation
• Available Bandwidth
• Hidden Markov Model
• Estimation Tool Traceband
• Performance Evaluation
• Moving Average
• Conclusions

3
Motivation

• Network applications can take advantage of the
  available bandwidth
• Network management, transport layer and routing
  protocols, overlay networks, traffic engineering,
  call admission control, among others
• Current available bandwidth tools cannot be
  applied in all application domains.
• Long convergence times
• Low accuracy
• High overhead
• Reducing estimation time while maintaining
  accuracy and low intrusion is an open research
  question.

Cesar D. Guerrero and Miguel A. Labrador. On the
Applicability of Available Bandwidth Estimation
Techniques and Tools. Submitted to IEEE Computer
Magazine
4
Available BandwidthDefinition
Tight link

• The Available Bandwidth (AB) is the non-utilized
  capacity in the link
• AB in the path is the minimum of all available
  bandwidths, which is the AB in the tight link

ABlink Clink 1-ulink Clink (Cross
traffic)link
5
Available BandwidthPacket pair sampling
Probing Packets
?in
P2
P1
Tight link capacity Ct
Cross traffic

• Available bandwidth can be estimated from the
  packet pair dispersions

6
Hidden Markov ModelAvailable Bandwidth Markov
Model
HIDDEN
7
Hidden Markov ModelElements and initial values
A
A
A
X1
Xt-1
Xt
XT
????
????
B
B
B
B
?1
? t
? t-1
? T

• States N10
• S S1, S2, ,S10
• Observation symbols M10
• V 1, 2, , 10
• Transition Probability Matrix A

• Observation Probabilities B
• Initial state probabilities ?
• ? P( X1 Si ) , 1 i N

fixed
random
random
8
Estimation Tool Traceband
traceband_snd.c
traceband_rcv.c
Read tm / Traceband total running time /
do Send a train of O30 packet pairs /
O50 every 10 trains / while running_time lt
tm Send last_estimation signal
Read_HMM(N,M,B) Aonestep_random(N,N) Pirandom(
N) do Receive a train of O packet
pairs For each packet pair at time
t1,..,T rel_dispersiont ?t (?out- ?in) /
?in observt ?t ceil(Mabs(1-?t)) Obser
v_seq O ?1, ?2, , ?T / Update A, Pi
/ HMMBWelch(HMM, O) / Find the state
sequence / QViterbi(HMM,O) Print ABmean(Q)
/ AB estimation / until last_estimation
signal
Sender
Receiver
Available at http//www.csee.usf.edu/guerrerc/t
raceband/soft.htm
9
Performance EvaluationTestbed and Experiments
Available Bandwidth 7 Mbps
10
Performance EvaluationPoisson Cross Traffic
Pathload
Spruce
Traceband
95 Confidence
11
Moving Average
i 5, R
Si-4,i(AB), meani-4,i(AB)
AB_MAi-4 meani-4,i(AB)
7,080,613
7,080,613
?AB(Si-4,iQ5,0.95)/(5)0.5
5,950,000 6,675,000 5,975,000 6,775,000 7,450,000
() 735,613
AB_MA(1) 6,345,000
(-) 735,613
no
7,450,000
ABi- meani-4,i(AB) gt ?AB
5,609,387
gt
yes
UPDATE
ABi meani-4,i(AB) ?AB
7,080,613
12
Performance EvaluationPoisson Cross Traffic
Moving Average
Pathload
Spruce
Traceband
95 Confidence
13
Performance EvaluationBursty Cross Traffic
Pathload
Spruce
Traceband
14
Conclusions

• Main contributions
• A novel HMM approach to estimate available
  bandwidth in an end to end path.
• Traceband A new available bandwidth estimation
  tool that can be used by applications requiring
  accurate, low overhead, and fast estimations.
• A Moving Average algorithm to filter peak
  values.
• Regarding to the performance evaluation
• Traceband is as accurate as Spruce and Pathload
  but considerably faster, and introduce minimum
  overhead.
• Traceband's convergence time is demonstrated
  using bursty cross-traffic, as it is the only
  tool that accurately reacts to zero-traffic
  periods

15
Thanks!
Questions?