CapProbe: An Efficient and Accurate Capacity Estimation Technique

1
CapProbe An Efficient and Accurate Capacity
Estimation Technique

• Rohit Kapoor, Ling-Jyh Chen, Li Lao, M.Y.
  Sanadidi, Mario Gerla
• Qualcomm Corp RD
• UCLA Computer Science Department

2
The Capacity Estimation Problem

• Estimate minimum link capacity on an Internet
  path, as seen at the IP level
• Design Goals
• End-to-end assume no help from routers
• Inexpensive Minimal additional traffic and
  processing
• Fast converges to capacity fast enough for the
  application

3
Applications

• Adaptive multimedia streaming
• Congestion control
• Capacity planning by ISPs
• Overlay network structuring
• Wireless link monitoring and mobility detection

4
Packet Pair Dispersion
5
Ideal Packet Dispersion

• No cross-traffic

Capacity (Packet Size) / (Dispersion)
6
Expansion of Dispersion

• Cross-traffic (CT) serviced between PP packets
• Second packet queues due to Cross Traffic (CT )gt
  expansion of dispersion gtUnder-estimation
• More pronounced when CT pkt size lt probe pkt size

7
Compression of Dispersion

• First packet queueing gt compressed dispersion gt
  Over-estimation
• More pronounced when CT pkt size gt probe pkt size

8
Previous Work

• Jacobsons Pathchar
• Estimates capacity for every link
• Sends varying size packets
• Relies on round trip delays
• Packet Pairs (PP)
• Crovella
• Capacity is reflected by the packet pair
  dispersion that occurs with highest frequency
• Lai
• Filters samples whose dispersion reflects a
  capacity greater than their potential bandwidth
• Both these techniques assume unimodal
  distribution
• Paxson showed distribution can be multimodal

9
Previous Work

• Dovrolis Work
• Analyzed under/over estimation of capacity
• Designed Pathrate
• First send packet pairs
• If multimodal, send packet trains
• Identifies modes to distinguish ADR (Asymptotic
  Dispersion Rate), PNCM (Post Narrow Capacity
  Mode) and Capacity Modes
• Previously proposed techniques have relied either
  on dispersion or delay

10
Key Observation

• First packet queues more than the second
• Compression
• Over-estimation
• Second packet queues more than the first
• Expansion
• Under-estimation
• Both expansion and compression are the result of
  probe packets experiencing queuing
• Sum of PP delay includes queuing delay

11
CapProbe Approach

• Filter PP samples that do not have minimum
  queuing time
• Dispersion of PP sample with minimum delay sum
  reflects capacity
• CapProbe combines both dispersion and e2e transit
  delay information

12
Techniques for Convergence Detection

• Consider set of packet pair probes 1n
• If min(d1) min(d2) ? min(d1d2), dispersion
  obtained from min delay sum may be distorted
• Above condition increases correct detection
  probability to that of a single packet (as
  opposed to packet pair)
• If above minimum delay sum condition is not
  satisfied in a run
• New run, with packet size of probes
• Increased if bandwidth estimated varied a lot
  across probes
• Errors in dispersion measured by OS
• Decreased if bandwidth estimated varied little
  across probes
• Packet sizes too large to go through without
  queuing

13
Experiments

• Simulations
• TCP (responsive), CBR (non-responsive), LRD
  (Pareto) cross-traffic
• Path-persistent, non-persistent cross-traffic

14
Simulations

• 6-hop path capacities 10, 7.5, 5.5, 4, 6, 8
  Mbps
• PP pkt size 200 bytes, CT pkt size 1000 bytes
• Path-Persistent TCP Cross-Traffic

15
Simulations

• PP pkt size CT pkt size 500 bytes
• Non-Persistent TCP Cross-Traffic

16
Simulations

• Non-Persistent UDP CBR Cross-Traffic
• Case where CapProbe may not work
• UDP (non-responsive), extremely intensive
• No correct samples are obtained

17
CapProbe Accuracy

• Sufficient requirement
• At least one PP sample where both packets
  experience no CT induced queuing delay.
• How realistic is this requirement?
• Internet is reactive (mostly TCP) high chance of
  some probing samples not being queued
• To validate, we performed extensive experiments
• Only cases where such undistorted samples are not
  obtained is when cross-traffic is UDP and very
  intensive (typically gt75 load)

18
Probability of Obtaining Sample

• Assuming PP samples arrive in a Poisson manner
• Poisson cross-traffic product of probabilities
• No queue in front of first packet p(0) 1 ?/µ
  
• No CT packets enter between the two packets
  (conservative estimate)
• Only dependent on arrival process
• p p(0) e- ?L/µ (1 ?/µ) e- ?L/µ
• Analysis also for Deterministic and Pareto
  cross-traffic

19
Probability of Obtaining Sample (cont)
Avg number of samples required to obtain an
unqueued PP for a single link Poisson
cross-traffic
Avg number of samples required to obtain an
unqueued PP for a single link LRD cross-traffic
20
Effect of Packet Size on Accuracy

• For CapProbe to estimate accurately
• Neither packet of the PP should queue due to
  cross traffic
• Second packet of PP
• Smaller ? less chances of queuing due to
  cross-traffic
• First packet of PP
• Probability of queuing independent of size
  (queuing theory)
• Thus, smaller PP packets ? higher probability of
  sample not subject to queuing
• Previous authors (Dovrolis) have shown that
• Smaller packets reduce chances of
  under-estimation but increase chances of
  over-estimation

21
Effect of Packet Size on Accuracy

• Our observations are entirely consistent with
  earlier ones
• For the second packet, smaller packet size ?
  Smaller probability of being queued ? Relative
  probability of queuing of first packet is
  increased ? Chances of over-estimation are
  increased

Frequency of occurrence of bandwidth samples when
packet size of probes is (a) 100 and (b) 1500
bytes
22
Measurements- Internet, Internet2 (Abilene),
Wireless (802.11, Bluetooth)

• CapProbe implemented using PING packets, sent in
  pairs

23
Issues

• CapProbe may be implemented either in the kernel
  or user mode
• Kernel mode more accurate, particularly over
  high-speed links
• One-way or round-trip estimation
• One-way
• Requires cooperation from receiver
• Can be used to estimate forward/reverse link
• Active vs passive
• Probing packets or data packets used as probes
• Heavy cross-traffic/extremely fast links
• Difficulty in correct estimation

24
Summary

• CapProbe is accurate, fast, and inexpensive,
  across a wide range of scenarios
• Potential applications in overlay structuring,
  and in case of fast changing wireless link speeds
• High-speed dispersion measurements needs more
  investigation
• CapProbe website http//nrl.cs.ucla.edu/CapProbe