The KaZaA Overlay: A Measurement Study

1
The KaZaA Overlay A Measurement Study

• ?a???s?as?
• ?e???ß??? F?te???
• Sta??p????? ???e??a
• F?t?p????? ?as????
• Authors Jian Liang, Rakesh Kumar, Keith W. Ross

2
Introduction

• ?? KaZaA e??a? µ?a ap? t?? s?µa?t???te?e?
  efa?µ???? t?? Internet ?s?? af??? t?? a???µ? t??
  ??µß?? p?? s?µµet????? a??? ?a? t? ßa?µ? t??
  ?????f???a?
• ???a p???µata e??a? ???st? ?s?? af??? t? overlay
  t?? KaZaA, t?? d?at???s? t?? overlay a??? ?a? t?
  signaling protocol

3
Introduction

• ?e pe??aµat??? ap?te??sµata p??sp???sa? ?a
  ?ata??ß??? p?? ?e?t????e?
• ?? overlay d??t??
• ? µ??a??sµ?? a?a??t?s??
• ? d?a?e???s? t?? index t?? s?st?µat??
• Signaling traffic

4

Overview of the KaZaA

• ??? ?at?????e? peers
• Super Nodes (SN)
• Ordinary Nodes (ON)
• ?? KaZaA peers d?af????? se availability,
  bandwidth, connectivity, CPU power ?a? NATed
  access

5

Overview of the KaZaA

• ???e ?? ?ta? e?s???e? st? KaZaA
• ??a???e? ?? pat??a t?? ??a? SN µe t?? ?p???
  d?at??e? µ?a ?µ?-µ???µ? TCP s??des?
• ???e? upload st?? pat??a t?? ta metadata t??
  a??e??? p?? ?a d?aµ???aste? st? KaZaA
• ???e SN d?at??e? ??a local index µe ????? t???
  ON p?? e??a? pa?d?? t??
• ?a???a? SN de? e??a? dedicated server

6

Overview of the KaZaA

• ?a metadata e??? a??e??? t?? ON pe??????? File
  name, File size, ContentHash, File descriptor
• File descriptor ???s?µ?p??e?ta? ??a a?t?st????s?
  keyword ?at? t?? d????e?a µ?a? e??t?s??
• ContentHash ?? KaZaA ?ata?e?µat??e? ???e a??e??
  se µ?a ?p???af?, ? ?p??a ???eta? t? ContentHash
  t?? a??e???

7

KaZaAs 2-tiered Overlay Network
ON
ON
ON
ON
ON
SN3
SN2
ON
ON
SN1
SN4
ON
ON
ON
ON
ON
ON
ON
ON
8
Topology Structure
ON to SN 100 - 160 s??d?se??? epe?d? ?p??????
3M ??µß?? a?? µ??a,????µe 30,000 SNs SN
to SN 30 50 s??d?se??? ???e SN e??a?
s??dedeµ???? µe t? 0.1 t?? s????????
a???µ?? t?? SNs
9
Topology Dynamics

• O µ?s?? ???? t?? d????e?a? t?? s??d?se?? ON-SN
  e??a? 34min ?a? SN-SN 11min
• ?a?at??????e p?? t? 32 t?? s??d?se?? SN-SN ?a?
  t? 38 t?? s??d?se?? ON-SN d?a????? ????te?? ap?
  30 seconds
• ? µ?s? d????e?a t?? s??d?se??, µe ?????
  µe?a??te?? ap? 30sec, e??a? 57min ??a ??-SN ?a?
  23min ??a SN-SN

10
Neighbor Selection

• ?ta? ??a? ON d?a???e? SN ap? t?? SN cash list
  ?aµß??e? ?p???? t?? locality ?a? t? workload t??
  SN
• Locality ?aµß??e? ?p??? t? RTT ?a? t?? IP
  prefixes
• 40 t?? ON-SN s??d?se?? ????? RTT lt 5 msec
• 60 t?? SN-SN s??d?se?? ????? RTT lt 50 msec
• RTT E. US ? Europe 100 msec

11
?????? ???? Supernode

• ? µ?s?? ?????? ???? e??? SN st? KaZaA e??a? 2.5
  ??e?

12
?p?f??? t?? Firewall

• ?????? default port 1214
• ?? administrator e????a d?aµ??f??a? ta firewalls
  ??a ap?t??p? t?? s??d?se??
• ???a d??aµ??? port number
• ??s???a µp???????ta? ?? s??d?se??
• ???? t? 3,6 ???s?µ?p??e? t? 1214 port, t? 96,3
  ???s?µ?p??e? ??a t??a?? port ap? 1024 ??? 65535
  ?a? e????st?? ???s?µ?p????? t? 80 port

13
?atast?at???s? t?? NAT

• ?? 30 t?? KaZaA peers e??a? p?s? ap? NATs
• ???ß??µa a? ? ? ???e? ?a s??de?e? µe t?? ? p??
  ??e? NAT address, de? µp??e? ?a a?????p????e? µ?a
  TCP s??des?
• ?e???? ep???s? t?? p??ß??µat?? (connection
  reversal)
• ? ? a?t? ?a ste??e? t?? a?t?s? st?? ? t?? st???e?
  st?? pat??a SN t?? ?
• ? pat??a? SN t?? ? st???e? ??a µ???µa ?a? t??
  ??e? ?a s??de?e? ?µesa µe t?? ?

14
??a?e???s? t?? Index

• ??et????µe t?? ?ata??µ? t?? metadata p?? ?????ta?
  upload st??? SN ap? t?? s??d?se?? t?? ON
• ?? 13 t?? ?? e??a? ?pe????? ??a t? 80 t??
  s???????? upload metadata
• ?a metadata ???e ?? pa?d??? sß????ta? aµ?s??
  µ???? t? pa?d? ap?s??de?e? ap? t?? pat??a SN
• ?e? ?p???e? a?ta??a?? index p????f???a?, d??ad?
  t?? metadata p?? s????????a? ap? ta pa?d?a ??,
  µeta?? SN-SN

15
Basic Design Principles

• Distributed Design
• All of its nodes run on user peers
• Exploiting Heterogeneity
• Super Nodes (SNs) and Ordinary Nodes (ONs)
• Load Balancing
• Each SN has roughly the same degree
• Locality in Neighbor Selection
• IP prefix and short RTT
• Connection Shuffling
• A larger set of SNs can be visited
• Efficient gossiping algorithms
• SNs learns about other SNs in the network
• Firewall avoidance and NAT circumvention
• Dynamic port and connection reversal

16

• ???t?se?????